d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
89f74f8743
freebsd-skq/sys/dev/sf/if_sf.c

2736 lines
76 KiB
C
Raw Normal View History

/* -> /*- for license, minor formatting changes
2005-01-07 02:29:27 +00:00
/*-
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
* Copyright (c) 1997, 1998, 1999
* Bill Paul <wpaul@ctr.columbia.edu>. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Bill Paul.
* 4. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
Remove duplicate FBSDID's, move others to their right place.
2003-11-14 17:16:58 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/*
* Adaptec AIC-6915 "Starfire" PCI fast ethernet driver for FreeBSD.
Update the Starfire driver comments and man page to include the URL of the AIC-6915 Programmer's Manual which I finally found online at Adaptec's site.
1999-09-26 18:35:30 +00:00
* Programming manual is available from:
Correct URL of the programming manual.
2004-11-17 23:36:11 +00:00
* http://download.adaptec.com/pdfs/user_guides/aic6915_pg.pdf.
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
*
* Written by Bill Paul <wpaul@ctr.columbia.edu>
* Department of Electical Engineering
* Columbia University, New York City
*/
/*
* The Adaptec AIC-6915 "Starfire" is a 64-bit 10/100 PCI ethernet
* controller designed with flexibility and reducing CPU load in mind.
* The Starfire offers high and low priority buffer queues, a
* producer/consumer index mechanism and several different buffer
* queue and completion queue descriptor types. Any one of a number
* of different driver designs can be used, depending on system and
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
* OS requirements. This driver makes use of type2 transmit frame
* descriptors to take full advantage of fragmented packets buffers
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
* and two RX buffer queues prioritized on size (one queue for small
* frames that will fit into a single mbuf, another with full size
* mbuf clusters for everything else). The producer/consumer indexes
* and completion queues are also used.
*
* One downside to the Starfire has to do with alignment: buffer
* queues must be aligned on 256-byte boundaries, and receive buffers
* must be aligned on longword boundaries. The receive buffer alignment
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
* causes problems on the strict alignment architecture, where the
* packet payload should be longword aligned. There is no simple way
* around this.
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
*
* For receive filtering, the Starfire offers 16 perfect filter slots
* and a 512-bit hash table.
*
* The Starfire has no internal transceiver, relying instead on an
* external MII-based transceiver. Accessing registers on external
* PHYs is done through a special register map rather than with the
* usual bitbang MDIO method.
*
* Acesssing the registers on the Starfire is a little tricky. The
* Starfire has a 512K internal register space. When programmed for
* PCI memory mapped mode, the entire register space can be accessed
* directly. However in I/O space mode, only 256 bytes are directly
* mapped into PCI I/O space. The other registers can be accessed
* indirectly using the SF_INDIRECTIO_ADDR and SF_INDIRECTIO_DATA
* registers inside the 256-byte I/O window.
*/
- Don't pollute opt_global.h with DEVICE_POLLING and introduce opt_device_polling.h - Include opt_device_polling.h into appropriate files. - Embrace with HAVE_KERNEL_OPTION_HEADERS the include in the files that can be compiled as loadable modules. Reviewed by: bde
2005-10-05 10:09:17 +00:00
#ifdef HAVE_KERNEL_OPTION_HEADERS
#include "opt_device_polling.h"
#endif
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#include <sys/param.h>
#include <sys/systm.h>
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#include <sys/bus.h>
#include <sys/endian.h>
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#include <sys/kernel.h>
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/rman.h>
Add missing <sys/module.h> includes
2004-05-30 20:00:41 +00:00
#include <sys/module.h>
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#include <sys/socket.h>
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#include <sys/sockio.h>
#include <sys/sysctl.h>
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#include <net/bpf.h>
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#include <net/if.h>
The r48589 promised to remove implicit inclusion of if_var.h soon. Prepare to this event, adding if_var.h to files that do need it. Also, include all includes that now are included due to implicit pollution via if_var.h Sponsored by: Netflix Sponsored by: Nginx, Inc.
2013-10-26 17:58:36 +00:00
#include <net/if_var.h>
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#include <net/if_arp.h>
#include <net/ethernet.h>
#include <net/if_dl.h>
#include <net/if_media.h>
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
#include <net/if_types.h>
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#include <net/if_vlan_var.h>
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
#include <dev/mii/mii.h>
#include <dev/mii/miivar.h>
Prefer new location of pci include files (which have only been in the tree for two or more years now), except in a few places where there's code to be compatible with older versions of FreeBSD.
2003-08-22 07:20:27 +00:00
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#include <machine/bus.h>
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Forced commit to note that sf(4) was repocopied from sys/pci and modified for its new location.
2008-01-21 04:23:47 +00:00
#include <dev/sf/if_sfreg.h>
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#include <dev/sf/starfire_rx.h>
#include <dev/sf/starfire_tx.h>
/* "device miibus" required. See GENERIC if you get errors here. */
#include "miibus_if.h"
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
- Express hard dependencies on bus (pci, isa, pccard) and network layer (ether). - Don't abuse module names to facilitate ifconfig module loading; such abuse isn't really needed. (And if we do need type information associated with a module then we should make it explicit and not use hacks.)
2003-04-15 06:37:30 +00:00
MODULE_DEPEND(sf, pci, 1, 1, 1);
MODULE_DEPEND(sf, ether, 1, 1, 1);
Depend on miibus. Note that if_aue doesn't strictly depend on usb because it uses the method interface for calls rather than using internal symbols, and because it's a child driver of usb and therefore will not try and do anything unless the parent usb code is loaded at some point. if_aue does strictly depend on miibus as it will fail to link if it is missing.
2000-04-29 13:41:57 +00:00
MODULE_DEPEND(sf, miibus, 1, 1, 1);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#undef SF_GFP_DEBUG
#define SF_CSUM_FEATURES (CSUM_TCP | CSUM_UDP)
/* Define this to activate partial TCP/UDP checksum offload. */
#undef SF_PARTIAL_CSUM_SUPPORT
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
static struct sf_type sf_devs[] = {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
{ AD_VENDORID, AD_DEVICEID_STARFIRE, "Adaptec AIC-6915 10/100BaseTX",
AD_SUBSYSID_62011_REV0, "Adaptec ANA-62011 (rev 0) 10/100BaseTX" },
{ AD_VENDORID, AD_DEVICEID_STARFIRE, "Adaptec AIC-6915 10/100BaseTX",
AD_SUBSYSID_62011_REV1, "Adaptec ANA-62011 (rev 1) 10/100BaseTX" },
{ AD_VENDORID, AD_DEVICEID_STARFIRE, "Adaptec AIC-6915 10/100BaseTX",
AD_SUBSYSID_62022, "Adaptec ANA-62022 10/100BaseTX" },
{ AD_VENDORID, AD_DEVICEID_STARFIRE, "Adaptec AIC-6915 10/100BaseTX",
AD_SUBSYSID_62044_REV0, "Adaptec ANA-62044 (rev 0) 10/100BaseTX" },
{ AD_VENDORID, AD_DEVICEID_STARFIRE, "Adaptec AIC-6915 10/100BaseTX",
AD_SUBSYSID_62044_REV1, "Adaptec ANA-62044 (rev 1) 10/100BaseTX" },
{ AD_VENDORID, AD_DEVICEID_STARFIRE, "Adaptec AIC-6915 10/100BaseTX",
AD_SUBSYSID_62020, "Adaptec ANA-62020 10/100BaseFX" },
{ AD_VENDORID, AD_DEVICEID_STARFIRE, "Adaptec AIC-6915 10/100BaseTX",
AD_SUBSYSID_69011, "Adaptec ANA-69011 10/100BaseTX" },
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
};
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int sf_probe(device_t);
static int sf_attach(device_t);
static int sf_detach(device_t);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static int sf_shutdown(device_t);
static int sf_suspend(device_t);
static int sf_resume(device_t);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void sf_intr(void *);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static void sf_tick(void *);
static void sf_stats_update(struct sf_softc *);
#ifndef __NO_STRICT_ALIGNMENT
static __inline void sf_fixup_rx(struct mbuf *);
#endif
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int sf_rxeof(struct sf_softc *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void sf_txeof(struct sf_softc *);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static int sf_encap(struct sf_softc *, struct mbuf **);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void sf_start(struct ifnet *);
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
static void sf_start_locked(struct ifnet *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int sf_ioctl(struct ifnet *, u_long, caddr_t);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static void sf_download_fw(struct sf_softc *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void sf_init(void *);
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
static void sf_init_locked(struct sf_softc *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void sf_stop(struct sf_softc *);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static void sf_watchdog(struct sf_softc *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int sf_ifmedia_upd(struct ifnet *);
Introduce sf_ifmedia_upd_locked() and have driver reset PHY before switching to selected media. While here, set if_drv_flags before switching to selected media.
2012-02-23 05:23:21 +00:00
static int sf_ifmedia_upd_locked(struct ifnet *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void sf_ifmedia_sts(struct ifnet *, struct ifmediareq *);
static void sf_reset(struct sf_softc *);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static int sf_dma_alloc(struct sf_softc *);
static void sf_dma_free(struct sf_softc *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int sf_init_rx_ring(struct sf_softc *);
static void sf_init_tx_ring(struct sf_softc *);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static int sf_newbuf(struct sf_softc *, int);
static void sf_rxfilter(struct sf_softc *);
static int sf_setperf(struct sf_softc *, int, uint8_t *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int sf_sethash(struct sf_softc *, caddr_t, int);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#ifdef notdef
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static int sf_setvlan(struct sf_softc *, int, uint32_t);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#endif
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static uint8_t sf_read_eeprom(struct sf_softc *, int);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int sf_miibus_readreg(device_t, int, int);
static int sf_miibus_writereg(device_t, int, int, int);
static void sf_miibus_statchg(device_t);
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
#ifdef DEVICE_POLLING
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int sf_poll(struct ifnet *ifp, enum poll_cmd cmd, int count);
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#endif
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static uint32_t csr_read_4(struct sf_softc *, int);
static void csr_write_4(struct sf_softc *, int, uint32_t);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void sf_txthresh_adjust(struct sf_softc *);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static int sf_sysctl_stats(SYSCTL_HANDLER_ARGS);
static int sysctl_int_range(SYSCTL_HANDLER_ARGS, int, int);
static int sysctl_hw_sf_int_mod(SYSCTL_HANDLER_ARGS);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
static device_method_t sf_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, sf_probe),
DEVMETHOD(device_attach, sf_attach),
DEVMETHOD(device_detach, sf_detach),
DEVMETHOD(device_shutdown, sf_shutdown),
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
DEVMETHOD(device_suspend, sf_suspend),
DEVMETHOD(device_resume, sf_resume),
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
/* MII interface */
DEVMETHOD(miibus_readreg, sf_miibus_readreg),
DEVMETHOD(miibus_writereg, sf_miibus_writereg),
DEVMETHOD(miibus_statchg, sf_miibus_statchg),
- There's no need to overwrite the default device method with the default one. Interestingly, these are actually the default for quite some time (bus_generic_driver_added(9) since r52045 and bus_generic_print_child(9) since r52045) but even recently added device drivers do this unnecessarily. Discussed with: jhb, marcel - While at it, use DEVMETHOD_END. Discussed with: jhb - Also while at it, use __FBSDID.
2011-11-22 21:28:20 +00:00
DEVMETHOD_END
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
};
static driver_t sf_driver = {
Grrr. Okay, changing the devnames was a bad idea. Put them back the way they were.
1999-09-20 08:47:11 +00:00
"sf",
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_methods,
sizeof(struct sf_softc),
};
static devclass_t sf_devclass;
- Express hard dependencies on bus (pci, isa, pccard) and network layer (ether). - Don't abuse module names to facilitate ifconfig module loading; such abuse isn't really needed. (And if we do need type information associated with a module then we should make it explicit and not use hacks.)
2003-04-15 06:37:30 +00:00
DRIVER_MODULE(sf, pci, sf_driver, sf_devclass, 0, 0);
Un-do the changes to the DRIVER_MODULE() declarations in these drivers. This whole idea isn't going to work until somebody makes the bus/kld code smarter. The idea here is to change the module's internal name from "foo" to "if_foo" so that ifconfig can tell a network driver from a non-network one. However doing this doesn't work correctly no matter how you slice it. For everything to work, you have to change the name in both the driver_t struct and the DRIVER_MODULE() declaration. The problems are: - If you change the name in both places, then the kernel thinks that the device's name is now "if_foo", so you get things like: if_foo0: <FOO ethernet> irq foo at device foo on pcifoo if_foo0: Ethernet address: foo:foo:foo:foo:foo:foo This is bogus. Now the device name doesn't agree with the logical interface name. There's no reason for this, and it violates the principle of least astonishment. - If you leave the name in the driver_t struct as "foo" and only change the names in the DRIVER_MODULE() declaration to "if_foo" then attaching drivers to child devices doesn't work because the names don't agree. This breaks miibus: drivers that need to have miibuses and PHY drivers attached never get them. In other words: damned if you do, damned if you don't. This needs to be thought through some more. Since the drivers that use miibus are broken, I have to change these all back in order to make them work again. Yes this will stop ifconfig from being able to demand load driver modules. On the whole, I'd rather have that than having the drivers not work at all.
1999-09-20 19:06:45 +00:00
DRIVER_MODULE(miibus, sf, miibus_driver, miibus_devclass, 0, 0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#define SF_SETBIT(sc, reg, x) \
Be consistent about functions being static. Properly put macro args in (). Spotted by: FlexeLint.
2002-10-16 09:14:59 +00:00
csr_write_4(sc, reg, csr_read_4(sc, reg) | (x))
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
#define SF_CLRBIT(sc, reg, x) \
Be consistent about functions being static. Properly put macro args in (). Spotted by: FlexeLint.
2002-10-16 09:14:59 +00:00
csr_write_4(sc, reg, csr_read_4(sc, reg) & ~(x))
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static uint32_t
csr_read_4(struct sf_softc *sc, int reg)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint32_t val;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_restype == SYS_RES_MEMORY)
val = CSR_READ_4(sc, (reg + SF_RMAP_INTREG_BASE));
else {
CSR_WRITE_4(sc, SF_INDIRECTIO_ADDR, reg + SF_RMAP_INTREG_BASE);
val = CSR_READ_4(sc, SF_INDIRECTIO_DATA);
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (val);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static uint8_t
sf_read_eeprom(struct sf_softc *sc, int reg)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint8_t val;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
val = (csr_read_4(sc, SF_EEADDR_BASE +
(reg & 0xFFFFFFFC)) >> (8 * (reg & 3))) & 0xFF;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (val);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
csr_write_4(struct sf_softc *sc, int reg, uint32_t val)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_restype == SYS_RES_MEMORY)
CSR_WRITE_4(sc, (reg + SF_RMAP_INTREG_BASE), val);
else {
CSR_WRITE_4(sc, SF_INDIRECTIO_ADDR, reg + SF_RMAP_INTREG_BASE);
CSR_WRITE_4(sc, SF_INDIRECTIO_DATA, val);
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
/*
* Copy the address 'mac' into the perfect RX filter entry at
* offset 'idx.' The perfect filter only has 16 entries so do
* some sanity tests.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_setperf(struct sf_softc *sc, int idx, uint8_t *mac)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
if (idx < 0 || idx > SF_RXFILT_PERFECT_CNT)
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (EINVAL);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
if (mac == NULL)
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (EINVAL);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_RXFILT_PERFECT_BASE +
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
(idx * SF_RXFILT_PERFECT_SKIP) + 0, mac[5] | (mac[4] << 8));
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_RXFILT_PERFECT_BASE +
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
(idx * SF_RXFILT_PERFECT_SKIP) + 4, mac[3] | (mac[2] << 8));
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_RXFILT_PERFECT_BASE +
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
(idx * SF_RXFILT_PERFECT_SKIP) + 8, mac[1] | (mac[0] << 8));
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
/*
* Set the bit in the 512-bit hash table that corresponds to the
* specified mac address 'mac.' If 'prio' is nonzero, update the
* priority hash table instead of the filter hash table.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_sethash(struct sf_softc *sc, caddr_t mac, int prio)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint32_t h;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
if (mac == NULL)
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (EINVAL);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Replace handrolled CRC calculation with ether_crc32_[lb]e().
2004-06-09 14:34:04 +00:00
h = ether_crc32_be(mac, ETHER_ADDR_LEN) >> 23;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
if (prio) {
SF_SETBIT(sc, SF_RXFILT_HASH_BASE + SF_RXFILT_HASH_PRIOOFF +
(SF_RXFILT_HASH_SKIP * (h >> 4)), (1 << (h & 0xF)));
} else {
SF_SETBIT(sc, SF_RXFILT_HASH_BASE + SF_RXFILT_HASH_ADDROFF +
(SF_RXFILT_HASH_SKIP * (h >> 4)), (1 << (h & 0xF)));
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
#ifdef notdef
/*
* Set a VLAN tag in the receive filter.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_setvlan(struct sf_softc *sc, int idx, uint32_t vlan)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
if (idx < 0 || idx >> SF_RXFILT_HASH_CNT)
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (EINVAL);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_RXFILT_HASH_BASE +
(idx * SF_RXFILT_HASH_SKIP) + SF_RXFILT_HASH_VLANOFF, vlan);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
#endif
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_miibus_readreg(device_t dev, int phy, int reg)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
struct sf_softc *sc;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
int i;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint32_t val = 0;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
sc = device_get_softc(dev);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
for (i = 0; i < SF_TIMEOUT; i++) {
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
val = csr_read_4(sc, SF_PHY_REG(phy, reg));
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((val & SF_MII_DATAVALID) != 0)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
break;
}
if (i == SF_TIMEOUT)
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
val &= SF_MII_DATAPORT;
if (val == 0xffff)
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (val);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_miibus_writereg(device_t dev, int phy, int reg, int val)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
struct sf_softc *sc;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
int i;
int busy;
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
sc = device_get_softc(dev);
csr_write_4(sc, SF_PHY_REG(phy, reg), val);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
for (i = 0; i < SF_TIMEOUT; i++) {
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
busy = csr_read_4(sc, SF_PHY_REG(phy, reg));
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((busy & SF_MII_BUSY) == 0)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
break;
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (0);
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_miibus_statchg(device_t dev)
{
struct sf_softc *sc;
struct mii_data *mii;
struct ifnet *ifp;
uint32_t val;
Remove taskqueue based MII stat change handler. Driver does not need deferred link state change processing. While I'm here, do not report current link status if interface is not UP.
2012-02-23 06:13:12 +00:00
sc = device_get_softc(dev);
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
mii = device_get_softc(sc->sf_miibus);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ifp = sc->sf_ifp;
if (mii == NULL || ifp == NULL ||
Remove taskqueue based MII stat change handler. Driver does not need deferred link state change processing. While I'm here, do not report current link status if interface is not UP.
2012-02-23 06:13:12 +00:00
(ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return;
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
With r232015, sf(4) gets correct speed/duplex of established link. Add more strict speed check in sf_miibus_statchg() and do not touch MAC config registers when driver lost a link.
2012-02-23 06:35:18 +00:00
sc->sf_link = 0;
if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) ==
(IFM_ACTIVE | IFM_AVALID)) {
switch (IFM_SUBTYPE(mii->mii_media_active)) {
case IFM_10_T:
case IFM_100_TX:
case IFM_100_FX:
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc->sf_link = 1;
With r232015, sf(4) gets correct speed/duplex of established link. Add more strict speed check in sf_miibus_statchg() and do not touch MAC config registers when driver lost a link.
2012-02-23 06:35:18 +00:00
break;
}
}
if (sc->sf_link == 0)
return;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
val = csr_read_4(sc, SF_MACCFG_1);
val &= ~SF_MACCFG1_FULLDUPLEX;
val &= ~(SF_MACCFG1_RX_FLOWENB | SF_MACCFG1_TX_FLOWENB);
if ((IFM_OPTIONS(mii->mii_media_active) & IFM_FDX) != 0) {
val |= SF_MACCFG1_FULLDUPLEX;
Modify the Adaptec "starfire" driver to reset the PHY on the MII bus before selecting a mode. The Seeq PHY chips on the Adaptec cards that use the AIC-6915 controller seem to need it in order to get them to change modes correctly. This corrects a problem that I noticed where my ANA-62022 board failed to correctly program the full duplex bit in the macconfig1 register when the interface was brought up. Running ifconfig sf0 would mask this problem in some cases because polling the PHY status would cause the miibus code to notice that full duplex was now needed and the statchg callback would be invoked to configure the duplex setting. However it would still get it wrong other times. Also changed sf_miibus_statchg() to program the IPG register to match the duplex setting in accordance with Adaptec manual's recommendations (0x15 for full duplex, 0x11 for half duplex).
1999-12-05 20:02:45 +00:00
csr_write_4(sc, SF_BKTOBKIPG, SF_IPGT_FDX);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#ifdef notyet
/* Configure flow-control bits. */
if ((IFM_OPTIONS(sc->sc_mii->mii_media_active) &
IFM_ETH_RXPAUSE) != 0)
val |= SF_MACCFG1_RX_FLOWENB;
if ((IFM_OPTIONS(sc->sc_mii->mii_media_active) &
IFM_ETH_TXPAUSE) != 0)
val |= SF_MACCFG1_TX_FLOWENB;
#endif
} else
Modify the Adaptec "starfire" driver to reset the PHY on the MII bus before selecting a mode. The Seeq PHY chips on the Adaptec cards that use the AIC-6915 controller seem to need it in order to get them to change modes correctly. This corrects a problem that I noticed where my ANA-62022 board failed to correctly program the full duplex bit in the macconfig1 register when the interface was brought up. Running ifconfig sf0 would mask this problem in some cases because polling the PHY status would cause the miibus code to notice that full duplex was now needed and the statchg callback would be invoked to configure the duplex setting. However it would still get it wrong other times. Also changed sf_miibus_statchg() to program the IPG register to match the duplex setting in accordance with Adaptec manual's recommendations (0x15 for full duplex, 0x11 for half duplex).
1999-12-05 20:02:45 +00:00
csr_write_4(sc, SF_BKTOBKIPG, SF_IPGT_HDX);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Make sure to reset MAC to take changes effect. */
csr_write_4(sc, SF_MACCFG_1, val | SF_MACCFG1_SOFTRESET);
DELAY(1000);
csr_write_4(sc, SF_MACCFG_1, val);
val = csr_read_4(sc, SF_TIMER_CTL);
if (IFM_SUBTYPE(mii->mii_media_active) == IFM_100_TX)
val |= SF_TIMER_TIMES_TEN;
else
val &= ~SF_TIMER_TIMES_TEN;
csr_write_4(sc, SF_TIMER_CTL, val);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_rxfilter(struct sf_softc *sc)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct ifnet *ifp;
int i;
struct ifmultiaddr *ifma;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint8_t dummy[ETHER_ADDR_LEN] = { 0, 0, 0, 0, 0, 0 };
uint32_t rxfilt;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->sf_ifp;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* First zot all the existing filters. */
for (i = 1; i < SF_RXFILT_PERFECT_CNT; i++)
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_setperf(sc, i, dummy);
for (i = SF_RXFILT_HASH_BASE; i < (SF_RXFILT_HASH_MAX + 1);
i += sizeof(uint32_t))
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, i, 0);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
rxfilt = csr_read_4(sc, SF_RXFILT);
rxfilt &= ~(SF_RXFILT_PROMISC | SF_RXFILT_ALLMULTI | SF_RXFILT_BROAD);
if ((ifp->if_flags & IFF_BROADCAST) != 0)
rxfilt |= SF_RXFILT_BROAD;
if ((ifp->if_flags & IFF_ALLMULTI) != 0 ||
(ifp->if_flags & IFF_PROMISC) != 0) {
if ((ifp->if_flags & IFF_PROMISC) != 0)
rxfilt |= SF_RXFILT_PROMISC;
if ((ifp->if_flags & IFF_ALLMULTI) != 0)
rxfilt |= SF_RXFILT_ALLMULTI;
goto done;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Now program new ones. */
i = 1;
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_rlock(ifp);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
TAILQ_FOREACH_REVERSE(ifma, &ifp->if_multiaddrs, ifmultihead,
ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
/*
* Program the first 15 multicast groups
* into the perfect filter. For all others,
* use the hash table.
*/
if (i < SF_RXFILT_PERFECT_CNT) {
sf_setperf(sc, i,
LLADDR((struct sockaddr_dl *)ifma->ifma_addr));
i++;
continue;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_sethash(sc,
LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_runlock(ifp);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
done:
csr_write_4(sc, SF_RXFILT, rxfilt);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
/*
* Set media options.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_ifmedia_upd(struct ifnet *ifp)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct sf_softc *sc;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
int error;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
sc = ifp->if_softc;
SF_LOCK(sc);
Introduce sf_ifmedia_upd_locked() and have driver reset PHY before switching to selected media. While here, set if_drv_flags before switching to selected media.
2012-02-23 05:23:21 +00:00
error = sf_ifmedia_upd_locked(ifp);
SF_UNLOCK(sc);
return (error);
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Introduce sf_ifmedia_upd_locked() and have driver reset PHY before switching to selected media. While here, set if_drv_flags before switching to selected media.
2012-02-23 05:23:21 +00:00
static int
sf_ifmedia_upd_locked(struct ifnet *ifp)
{
struct sf_softc *sc;
struct mii_data *mii;
struct mii_softc *miisc;
sc = ifp->if_softc;
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
mii = device_get_softc(sc->sf_miibus);
- Remove attempts to implement setting of BMCR_LOOP/MIIF_NOLOOP (reporting IFM_LOOP based on BMCR_LOOP is left in place though as it might provide useful for debugging). For most mii(4) drivers it was unclear whether the PHYs driven by them actually support loopback or not. Moreover, typically loopback mode also needs to be activated on the MAC, which none of the Ethernet drivers using mii(4) implements. Given that loopback media has no real use (and obviously hardly had a chance to actually work) besides for driver development (which just loopback mode should be sufficient for though, i.e one doesn't necessary need support for loopback media) support for it is just dropped as both NetBSD and OpenBSD already did quite some time ago. - Let mii_phy_add_media() also announce the support of IFM_NONE. - Restructure the PHY entry points to use a structure of entry points instead of discrete function pointers, and extend this to include a "reset" entry point. Make sure any PHY-specific reset routine is always used, and provide one for lxtphy(4) which disables MII interrupts (as is done for a few other PHYs we have drivers for). This includes changing NIC drivers which previously just called the generic mii_phy_reset() to now actually call the PHY-specific reset routine, which might be crucial in some cases. While at it, the redundant checks in these NIC drivers for mii->mii_instance not being zero before calling the reset routines were removed because as soon as one PHY driver attaches mii->mii_instance is incremented and we hardly can end up in their media change callbacks etc if no PHY driver has attached as mii_attach() would have failed in that case and not attach a miibus(4) instance. Consequently, NIC drivers now no longer should call mii_phy_reset() directly, so it was removed from EXPORT_SYMS. - Add a mii_phy_dev_attach() as a companion helper to mii_phy_dev_probe(). The purpose of that function is to perform the common steps to attach a PHY driver instance and to hook it up to the miibus(4) instance and to optionally also handle the probing, addition and initialization of the supported media. So all a PHY driver without any special requirements has to do in its bus attach method is to call mii_phy_dev_attach() along with PHY-specific MIIF_* flags, a pointer to its PHY functions and the add_media set to one. All PHY drivers were updated to take advantage of mii_phy_dev_attach() as appropriate. Along with these changes the capability mask was added to the mii_softc structure so PHY drivers taking advantage of mii_phy_dev_attach() but still handling media on their own do not need to fiddle with the MII attach arguments anyway. - Keep track of the PHY offset in the mii_softc structure. This is done for compatibility with NetBSD/OpenBSD. - Keep track of the PHY's OUI, model and revision in the mii_softc structure. Several PHY drivers require this information also after attaching and previously had to wrap their own softc around mii_softc. NetBSD/OpenBSD also keep track of the model and revision on their mii_softc structure. All PHY drivers were updated to take advantage as appropriate. - Convert the mebers of the MII data structure to unsigned where appropriate. This is partly inspired by NetBSD/OpenBSD. - According to IEEE 802.3-2002 the bits actually have to be reversed when mapping an OUI to the MII ID registers. All PHY drivers and miidevs where changed as necessary. Actually this now again allows to largely share miidevs with NetBSD, which fixed this problem already 9 years ago. Consequently miidevs was synced as far as possible. - Add MIIF_NOMANPAUSE and mii_phy_flowstatus() calls to drivers that weren't explicitly converted to support flow control before. It's unclear whether flow control actually works with these but typically it should and their net behavior should be more correct with these changes in place than without if the MAC driver sets MIIF_DOPAUSE. Obtained from: NetBSD (partially) Reviewed by: yongari (earlier version), silence on arch@ and net@
2011-05-03 19:51:29 +00:00
LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
PHY_RESET(miisc);
Introduce sf_ifmedia_upd_locked() and have driver reset PHY before switching to selected media. While here, set if_drv_flags before switching to selected media.
2012-02-23 05:23:21 +00:00
return (mii_mediachg(mii));
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
/*
* Report current media status.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct sf_softc *sc;
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
struct mii_data *mii;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sc = ifp->if_softc;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK(sc);
Remove taskqueue based MII stat change handler. Driver does not need deferred link state change processing. While I'm here, do not report current link status if interface is not UP.
2012-02-23 06:13:12 +00:00
if ((ifp->if_flags & IFF_UP) == 0) {
SF_UNLOCK(sc);
return;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Remove taskqueue based MII stat change handler. Driver does not need deferred link state change processing. While I'm here, do not report current link status if interface is not UP.
2012-02-23 06:13:12 +00:00
mii = device_get_softc(sc->sf_miibus);
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
mii_pollstat(mii);
ifmr->ifm_active = mii->mii_media_active;
ifmr->ifm_status = mii->mii_media_status;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_UNLOCK(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_softc *sc;
struct ifreq *ifr;
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
struct mii_data *mii;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
int error, mask;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc = ifp->if_softc;
ifr = (struct ifreq *)data;
error = 0;
switch (command) {
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
case SIOCSIFFLAGS:
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
if (ifp->if_flags & IFF_UP) {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
if ((ifp->if_flags ^ sc->sf_if_flags) &
(IFF_PROMISC | IFF_ALLMULTI))
sf_rxfilter(sc);
} else {
if (sc->sf_detach == 0)
sf_init_locked(sc);
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
} else {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_stop(sc);
}
Modify the Adaptec "starfire" driver to reset the PHY on the MII bus before selecting a mode. The Seeq PHY chips on the Adaptec cards that use the AIC-6915 controller seem to need it in order to get them to change modes correctly. This corrects a problem that I noticed where my ANA-62022 board failed to correctly program the full duplex bit in the macconfig1 register when the interface was brought up. Running ifconfig sf0 would mask this problem in some cases because polling the PHY status would cause the miibus code to notice that full duplex was now needed and the statchg callback would be invoked to configure the duplex setting. However it would still get it wrong other times. Also changed sf_miibus_statchg() to program the IPG register to match the duplex setting in accordance with Adaptec manual's recommendations (0x15 for full duplex, 0x11 for half duplex).
1999-12-05 20:02:45 +00:00
sc->sf_if_flags = ifp->if_flags;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_UNLOCK(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK(sc);
No need to reprogram hardware RX filter when driver is not running.
2012-02-23 05:25:14 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
sf_rxfilter(sc);
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_UNLOCK(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
break;
case SIOCGIFMEDIA:
case SIOCSIFMEDIA:
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
mii = device_get_softc(sc->sf_miibus);
error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, command);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
break;
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
case SIOCSIFCAP:
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
mask = ifr->ifr_reqcap ^ ifp->if_capenable;
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#ifdef DEVICE_POLLING
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((mask & IFCAP_POLLING) != 0) {
if ((ifr->ifr_reqcap & IFCAP_POLLING) != 0) {
error = ether_poll_register(sf_poll, ifp);
if (error != 0)
break;
SF_LOCK(sc);
/* Disable interrupts. */
csr_write_4(sc, SF_IMR, 0);
ifp->if_capenable |= IFCAP_POLLING;
SF_UNLOCK(sc);
} else {
error = ether_poll_deregister(ifp);
/* Enable interrupts. */
SF_LOCK(sc);
csr_write_4(sc, SF_IMR, SF_INTRS);
ifp->if_capenable &= ~IFCAP_POLLING;
SF_UNLOCK(sc);
}
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
}
#endif /* DEVICE_POLLING */
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((mask & IFCAP_TXCSUM) != 0) {
if ((IFCAP_TXCSUM & ifp->if_capabilities) != 0) {
SF_LOCK(sc);
ifp->if_capenable ^= IFCAP_TXCSUM;
if ((IFCAP_TXCSUM & ifp->if_capenable) != 0) {
ifp->if_hwassist |= SF_CSUM_FEATURES;
SF_SETBIT(sc, SF_GEN_ETH_CTL,
SF_ETHCTL_TXGFP_ENB);
} else {
ifp->if_hwassist &= ~SF_CSUM_FEATURES;
SF_CLRBIT(sc, SF_GEN_ETH_CTL,
SF_ETHCTL_TXGFP_ENB);
}
SF_UNLOCK(sc);
}
}
if ((mask & IFCAP_RXCSUM) != 0) {
if ((IFCAP_RXCSUM & ifp->if_capabilities) != 0) {
SF_LOCK(sc);
ifp->if_capenable ^= IFCAP_RXCSUM;
if ((IFCAP_RXCSUM & ifp->if_capenable) != 0)
SF_SETBIT(sc, SF_GEN_ETH_CTL,
SF_ETHCTL_RXGFP_ENB);
else
SF_CLRBIT(sc, SF_GEN_ETH_CTL,
SF_ETHCTL_RXGFP_ENB);
SF_UNLOCK(sc);
}
}
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
break;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
default:
o track either_ifattach/ether_ifdetach API changes o use if_input for input packet processing o don't strip the Ethernet header for input packets o use BPF_* macros bpf tapping o call ether_ioctl to handle default ioctl case o track vlan changes Reviewed by: many Approved by: re
2002-11-14 23:49:09 +00:00
error = ether_ioctl(ifp, command, data);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
break;
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (error);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_reset(struct sf_softc *sc)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
int i;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_GEN_ETH_CTL, 0);
SF_SETBIT(sc, SF_MACCFG_1, SF_MACCFG1_SOFTRESET);
DELAY(1000);
SF_CLRBIT(sc, SF_MACCFG_1, SF_MACCFG1_SOFTRESET);
SF_SETBIT(sc, SF_PCI_DEVCFG, SF_PCIDEVCFG_RESET);
for (i = 0; i < SF_TIMEOUT; i++) {
DELAY(10);
if (!(csr_read_4(sc, SF_PCI_DEVCFG) & SF_PCIDEVCFG_RESET))
break;
}
if (i == SF_TIMEOUT)
Consistently use if_printf() only in interface methods: if_start, if_ioctl, if_watchdog, etc, or in functions that are used by these methods only. In all other cases use device_printf(). This also fixes several panics, when if_printf() is called before softc->ifp was initialized. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 10:40:54 +00:00
device_printf(sc->sf_dev, "reset never completed!\n");
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* Wait a little while for the chip to get its brains in order. */
DELAY(1000);
}
/*
* Probe for an Adaptec AIC-6915 chip. Check the PCI vendor and device
* IDs against our list and return a device name if we find a match.
* We also check the subsystem ID so that we can identify exactly which
* NIC has been found, if possible.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_probe(device_t dev)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct sf_type *t;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint16_t vid;
uint16_t did;
uint16_t sdid;
int i;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
vid = pci_get_vendor(dev);
did = pci_get_device(dev);
sdid = pci_get_subdevice(dev);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
t = sf_devs;
for (i = 0; i < sizeof(sf_devs) / sizeof(sf_devs[0]); i++, t++) {
if (vid == t->sf_vid && did == t->sf_did) {
if (sdid == t->sf_sdid) {
device_set_desc(dev, t->sf_sname);
Return BUS_PROBE_DEFAULT instead of 0.
2005-02-24 21:32:56 +00:00
return (BUS_PROBE_DEFAULT);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
}
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (vid == AD_VENDORID && did == AD_DEVICEID_STARFIRE) {
/* unkown subdevice */
device_set_desc(dev, sf_devs[0].sf_name);
return (BUS_PROBE_DEFAULT);
}
return (ENXIO);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
/*
* Attach the interface. Allocate softc structures, do ifmedia
* setup and ethernet/BPF attach.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_attach(device_t dev)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
int i;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
struct sf_softc *sc;
struct ifnet *ifp;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint32_t reg;
Use if_printf() and device_printf() and axe sf_unit from the softc as a result.
2005-08-10 20:24:39 +00:00
int rid, error = 0;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint8_t eaddr[ETHER_ADDR_LEN];
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sc = device_get_softc(dev);
Consistently use if_printf() only in interface methods: if_start, if_ioctl, if_watchdog, etc, or in functions that are used by these methods only. In all other cases use device_printf(). This also fixes several panics, when if_printf() is called before softc->ifp was initialized. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 10:40:54 +00:00
sc->sf_dev = dev;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Change callers of mtx_init() to pass in an appropriate lock type name. In most cases NULL is passed, but in some cases such as network driver locks (which use the MTX_NETWORK_LOCK macro) and UMA zone locks, a name is used. Tested on: i386, alpha, sparc64
2002-04-04 21:03:38 +00:00
mtx_init(&sc->sf_mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK,
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
MTX_DEF);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
callout_init_mtx(&sc->sf_co, &sc->sf_mtx, 0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/*
* Map control/status registers.
*/
Big round of minor updates: - Use pci_get_powerstate()/pci_set_powerstate() in all the other drivers that need them so we don't have to fiddle with the PCI power management registers directly. - Use pci_enable_busmaster()/pci_enable_io() to turn on busmastering and PIO/memory mapped accesses. - Add support to the RealTek driver for the D-Link DFE-530TX+ which has a RealTek 8139 with its own PCI ID. (Submitted by Jason Wright) - Have the SiS 900/National DP83815 driver be sure to disable PME mode in sis_reset(). This apparently fixes a problem on some motherboards where the DP83815 chip fails to receive packets. (Submitted by Chuck McCrobie <mccrobie@cablespeed.com>)
2001-02-21 20:54:22 +00:00
pci_enable_busmaster(dev);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* Prefer memory space register mapping over I/O space as the
* hardware requires lots of register access to get various
* producer/consumer index during Tx/Rx operation. However this
* requires large memory space(512K) to map the entire register
* space.
*/
sc->sf_rid = PCIR_BAR(0);
sc->sf_restype = SYS_RES_MEMORY;
sc->sf_res = bus_alloc_resource_any(dev, sc->sf_restype, &sc->sf_rid,
RF_ACTIVE);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
if (sc->sf_res == NULL) {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
reg = pci_read_config(dev, PCIR_BAR(0), 4);
if ((reg & PCIM_BAR_MEM_64) == PCIM_BAR_MEM_64)
sc->sf_rid = PCIR_BAR(2);
else
sc->sf_rid = PCIR_BAR(1);
sc->sf_restype = SYS_RES_IOPORT;
sc->sf_res = bus_alloc_resource_any(dev, sc->sf_restype,
&sc->sf_rid, RF_ACTIVE);
if (sc->sf_res == NULL) {
device_printf(dev, "couldn't allocate resources\n");
mtx_destroy(&sc->sf_mtx);
return (ENXIO);
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (bootverbose)
device_printf(dev, "using %s space register mapping\n",
sc->sf_restype == SYS_RES_MEMORY ? "memory" : "I/O");
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
reg = pci_read_config(dev, PCIR_CACHELNSZ, 1);
if (reg == 0) {
/*
* If cache line size is 0, MWI is not used at all, so set
* reasonable default. AIC-6915 supports 0, 4, 8, 16, 32
* and 64.
*/
reg = 16;
device_printf(dev, "setting PCI cache line size to %u\n", reg);
pci_write_config(dev, PCIR_CACHELNSZ, reg, 1);
} else {
if (bootverbose)
device_printf(dev, "PCI cache line size : %u\n", reg);
}
/* Enable MWI. */
reg = pci_read_config(dev, PCIR_COMMAND, 2);
reg |= PCIM_CMD_MWRICEN;
pci_write_config(dev, PCIR_COMMAND, reg, 2);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Allocate interrupt. */
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
rid = 0;
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->sf_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
RF_SHAREABLE | RF_ACTIVE);
if (sc->sf_irq == NULL) {
Use if_printf() and device_printf() and axe sf_unit from the softc as a result.
2005-08-10 20:24:39 +00:00
device_printf(dev, "couldn't map interrupt\n");
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
error = ENXIO;
goto fail;
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "stats", CTLTYPE_INT | CTLFLAG_RW, sc, 0,
sf_sysctl_stats, "I", "Statistics");
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "int_mod", CTLTYPE_INT | CTLFLAG_RW,
&sc->sf_int_mod, 0, sysctl_hw_sf_int_mod, "I",
"sf interrupt moderation");
/* Pull in device tunables. */
sc->sf_int_mod = SF_IM_DEFAULT;
error = resource_int_value(device_get_name(dev), device_get_unit(dev),
"int_mod", &sc->sf_int_mod);
if (error == 0) {
if (sc->sf_int_mod < SF_IM_MIN ||
sc->sf_int_mod > SF_IM_MAX) {
device_printf(dev, "int_mod value out of range; "
"using default: %d\n", SF_IM_DEFAULT);
sc->sf_int_mod = SF_IM_DEFAULT;
}
}
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* Reset the adapter. */
sf_reset(sc);
/*
* Get station address from the EEPROM.
*/
for (i = 0; i < ETHER_ADDR_LEN; i++)
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
eaddr[i] =
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_read_eeprom(sc, SF_EE_NODEADDR + ETHER_ADDR_LEN - i);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Allocate DMA resources. */
if (sf_dma_alloc(sc) != 0) {
error = ENOSPC;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
goto fail;
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc->sf_txthresh = SF_MIN_TX_THRESHOLD;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Move if_alloc() up so it's before mii_phy_probe().
2005-06-11 01:37:46 +00:00
ifp = sc->sf_ifp = if_alloc(IFT_ETHER);
if (ifp == NULL) {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
device_printf(dev, "can not allocate ifnet structure\n");
Move if_alloc() up so it's before mii_phy_probe().
2005-06-11 01:37:46 +00:00
error = ENOSPC;
goto fail;
}
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
/* Do MII setup. */
Converted the remainder of the NIC drivers to use the mii_attach() introduced in r213878 instead of mii_phy_probe(). Unlike r213893 these are only straight forward conversions though. Reviewed by: yongari
2010-10-15 15:00:30 +00:00
error = mii_attach(dev, &sc->sf_miibus, ifp, sf_ifmedia_upd,
sf_ifmedia_sts, BMSR_DEFCAPMASK, MII_PHY_ANY, MII_OFFSET_ANY, 0);
if (error != 0) {
device_printf(dev, "attaching PHYs failed\n");
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
goto fail;
}
ifp->if_softc = sc;
Replace the if_name and if_unit members of struct ifnet with new members if_xname, if_dname, and if_dunit. if_xname is the name of the interface and if_dname/unit are the driver name and instance. This change paves the way for interface renaming and enhanced pseudo device creation and configuration symantics. Approved By: re (in principle) Reviewed By: njl, imp Tested On: i386, amd64, sparc64 Obtained From: NetBSD (if_xname)
2003-10-31 18:32:15 +00:00
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
ifp->if_ioctl = sf_ioctl;
ifp->if_start = sf_start;
ifp->if_init = sf_init;
Add altq support. Patch by mlaier. Approved by: mlaier MFC after: 2 weeks
2004-11-12 18:12:04 +00:00
IFQ_SET_MAXLEN(&ifp->if_snd, SF_TX_DLIST_CNT - 1);
ifp->if_snd.ifq_drv_maxlen = SF_TX_DLIST_CNT - 1;
IFQ_SET_READY(&ifp->if_snd);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* With the help of firmware, AIC-6915 supports
* Tx/Rx TCP/UDP checksum offload.
*/
ifp->if_hwassist = SF_CSUM_FEATURES;
ifp->if_capabilities = IFCAP_HWCSUM;
/*
* Call MI attach routine.
*/
ether_ifattach(ifp, eaddr);
/* VLAN capability setup. */
ifp->if_capabilities |= IFCAP_VLAN_MTU;
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
ifp->if_capenable = ifp->if_capabilities;
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
#ifdef DEVICE_POLLING
ifp->if_capabilities |= IFCAP_POLLING;
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#endif
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/*
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
* Tell the upper layer(s) we support long frames.
* Must appear after the call to ether_ifattach() because
* ether_ifattach() sets ifi_hdrlen to the default value.
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
*/
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
/* Hook interrupt last to avoid having to lock softc */
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
error = bus_setup_intr(dev, sc->sf_irq, INTR_TYPE_NET | INTR_MPSAFE,
o break newbus api: add a new argument of type driver_filter_t to bus_setup_intr() o add an int return code to all fast handlers o retire INTR_FAST/IH_FAST For more info: http://docs.freebsd.org/cgi/getmsg.cgi?fetch=465712+0+current/freebsd-current Reviewed by: many Approved by: re@
2007-02-23 12:19:07 +00:00
NULL, sf_intr, sc, &sc->sf_intrhand);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
if (error) {
Use if_printf() and device_printf() and axe sf_unit from the softc as a result.
2005-08-10 20:24:39 +00:00
device_printf(dev, "couldn't set up irq\n");
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
ether_ifdetach(ifp);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
fail:
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
if (error)
sf_detach(dev);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (error);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
/*
* Shutdown hardware and free up resources. This can be called any
* time after the mutex has been initialized. It is called in both
* the error case in attach and the normal detach case so it needs
* to be careful about only freeing resources that have actually been
* allocated.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_detach(device_t dev)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct sf_softc *sc;
struct ifnet *ifp;
sc = device_get_softc(dev);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->sf_ifp;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#ifdef DEVICE_POLLING
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (ifp != NULL && ifp->if_capenable & IFCAP_POLLING)
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
ether_poll_deregister(ifp);
#endif
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
/* These should only be active if attach succeeded */
Use newly minted device_is_attached rather than device_is_alive to see if attach succeeded. device_is_alive just tells us that probe succeeded. Since we were using it to do things like detach net interfaces, this caused problems when there were errors in the attach routine. Symptoms of problem reported by: martin blapp
2003-04-21 18:34:04 +00:00
if (device_is_attached(dev)) {
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc->sf_detach = 1;
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
sf_stop(sc);
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_UNLOCK(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
callout_drain(&sc->sf_co);
if (ifp != NULL)
ether_ifdetach(ifp);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_miibus) {
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
device_delete_child(dev, sc->sf_miibus);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc->sf_miibus = NULL;
}
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
bus_generic_detach(dev);
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_intrhand != NULL)
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
bus_teardown_intr(dev, sc->sf_irq, sc->sf_intrhand);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_irq != NULL)
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
bus_release_resource(dev, SYS_RES_IRQ, 0, sc->sf_irq);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_res != NULL)
bus_release_resource(dev, sc->sf_restype, sc->sf_rid,
sc->sf_res);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_dma_free(sc);
if (ifp != NULL)
In detach method, move if_free() after bus_teardown_intr().
2005-10-13 21:11:20 +00:00
if_free(ifp);
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
mtx_destroy(&sc->sf_mtx);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_dmamap_arg {
bus_addr_t sf_busaddr;
};
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_dmamap_arg *ctx;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (error != 0)
return;
ctx = arg;
ctx->sf_busaddr = segs[0].ds_addr;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_dma_alloc(struct sf_softc *sc)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_dmamap_arg ctx;
struct sf_txdesc *txd;
struct sf_rxdesc *rxd;
bus_addr_t lowaddr;
bus_addr_t rx_ring_end, rx_cring_end;
bus_addr_t tx_ring_end, tx_cring_end;
int error, i;
lowaddr = BUS_SPACE_MAXADDR;
again:
/* Create parent DMA tag. */
error = bus_dma_tag_create(
bus_get_dma_tag(sc->sf_dev), /* parent */
1, 0, /* alignment, boundary */
lowaddr, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
BUS_SPACE_MAXSIZE_32BIT, /* maxsize */
0, /* nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->sf_cdata.sf_parent_tag);
if (error != 0) {
device_printf(sc->sf_dev, "failed to create parent DMA tag\n");
goto fail;
}
/* Create tag for Tx ring. */
error = bus_dma_tag_create(sc->sf_cdata.sf_parent_tag,/* parent */
SF_RING_ALIGN, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
SF_TX_DLIST_SIZE, /* maxsize */
1, /* nsegments */
SF_TX_DLIST_SIZE, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->sf_cdata.sf_tx_ring_tag);
if (error != 0) {
device_printf(sc->sf_dev, "failed to create Tx ring DMA tag\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Create tag for Tx completion ring. */
error = bus_dma_tag_create(sc->sf_cdata.sf_parent_tag,/* parent */
SF_RING_ALIGN, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
SF_TX_CLIST_SIZE, /* maxsize */
1, /* nsegments */
SF_TX_CLIST_SIZE, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->sf_cdata.sf_tx_cring_tag);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to create Tx completion ring DMA tag\n");
goto fail;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Create tag for Rx ring. */
error = bus_dma_tag_create(sc->sf_cdata.sf_parent_tag,/* parent */
SF_RING_ALIGN, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
SF_RX_DLIST_SIZE, /* maxsize */
1, /* nsegments */
SF_RX_DLIST_SIZE, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->sf_cdata.sf_rx_ring_tag);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to create Rx ring DMA tag\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Create tag for Rx completion ring. */
error = bus_dma_tag_create(sc->sf_cdata.sf_parent_tag,/* parent */
SF_RING_ALIGN, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
SF_RX_CLIST_SIZE, /* maxsize */
1, /* nsegments */
SF_RX_CLIST_SIZE, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->sf_cdata.sf_rx_cring_tag);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to create Rx completion ring DMA tag\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Create tag for Tx buffers. */
error = bus_dma_tag_create(sc->sf_cdata.sf_parent_tag,/* parent */
1, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
MCLBYTES * SF_MAXTXSEGS, /* maxsize */
SF_MAXTXSEGS, /* nsegments */
MCLBYTES, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->sf_cdata.sf_tx_tag);
if (error != 0) {
device_printf(sc->sf_dev, "failed to create Tx DMA tag\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Create tag for Rx buffers. */
error = bus_dma_tag_create(sc->sf_cdata.sf_parent_tag,/* parent */
SF_RX_ALIGN, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
MCLBYTES, /* maxsize */
1, /* nsegments */
MCLBYTES, /* maxsegsize */
0, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->sf_cdata.sf_rx_tag);
if (error != 0) {
device_printf(sc->sf_dev, "failed to create Rx DMA tag\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Allocate DMA'able memory and load the DMA map for Tx ring. */
error = bus_dmamem_alloc(sc->sf_cdata.sf_tx_ring_tag,
(void **)&sc->sf_rdata.sf_tx_ring, BUS_DMA_WAITOK |
BUS_DMA_COHERENT | BUS_DMA_ZERO, &sc->sf_cdata.sf_tx_ring_map);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to allocate DMA'able memory for Tx ring\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ctx.sf_busaddr = 0;
error = bus_dmamap_load(sc->sf_cdata.sf_tx_ring_tag,
sc->sf_cdata.sf_tx_ring_map, sc->sf_rdata.sf_tx_ring,
SF_TX_DLIST_SIZE, sf_dmamap_cb, &ctx, 0);
if (error != 0 || ctx.sf_busaddr == 0) {
device_printf(sc->sf_dev,
"failed to load DMA'able memory for Tx ring\n");
goto fail;
}
sc->sf_rdata.sf_tx_ring_paddr = ctx.sf_busaddr;
Drop the driver lock around calls to if_input to avoid a LOR when the packets are immediately returned for sending (e.g. when bridging or packet forwarding). There are more efficient ways to do this but for now use the least intrusive approach. Reviewed by: imp, rwatson
2003-11-14 19:00:32 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* Allocate DMA'able memory and load the DMA map for Tx completion ring.
*/
error = bus_dmamem_alloc(sc->sf_cdata.sf_tx_cring_tag,
(void **)&sc->sf_rdata.sf_tx_cring, BUS_DMA_WAITOK |
BUS_DMA_COHERENT | BUS_DMA_ZERO, &sc->sf_cdata.sf_tx_cring_map);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to allocate DMA'able memory for "
"Tx completion ring\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ctx.sf_busaddr = 0;
error = bus_dmamap_load(sc->sf_cdata.sf_tx_cring_tag,
sc->sf_cdata.sf_tx_cring_map, sc->sf_rdata.sf_tx_cring,
SF_TX_CLIST_SIZE, sf_dmamap_cb, &ctx, 0);
if (error != 0 || ctx.sf_busaddr == 0) {
device_printf(sc->sf_dev,
"failed to load DMA'able memory for Tx completion ring\n");
goto fail;
}
sc->sf_rdata.sf_tx_cring_paddr = ctx.sf_busaddr;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Allocate DMA'able memory and load the DMA map for Rx ring. */
error = bus_dmamem_alloc(sc->sf_cdata.sf_rx_ring_tag,
(void **)&sc->sf_rdata.sf_rx_ring, BUS_DMA_WAITOK |
BUS_DMA_COHERENT | BUS_DMA_ZERO, &sc->sf_cdata.sf_rx_ring_map);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to allocate DMA'able memory for Rx ring\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ctx.sf_busaddr = 0;
error = bus_dmamap_load(sc->sf_cdata.sf_rx_ring_tag,
sc->sf_cdata.sf_rx_ring_map, sc->sf_rdata.sf_rx_ring,
SF_RX_DLIST_SIZE, sf_dmamap_cb, &ctx, 0);
if (error != 0 || ctx.sf_busaddr == 0) {
device_printf(sc->sf_dev,
"failed to load DMA'able memory for Rx ring\n");
goto fail;
}
sc->sf_rdata.sf_rx_ring_paddr = ctx.sf_busaddr;
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* Allocate DMA'able memory and load the DMA map for Rx completion ring.
*/
error = bus_dmamem_alloc(sc->sf_cdata.sf_rx_cring_tag,
(void **)&sc->sf_rdata.sf_rx_cring, BUS_DMA_WAITOK |
BUS_DMA_COHERENT | BUS_DMA_ZERO, &sc->sf_cdata.sf_rx_cring_map);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to allocate DMA'able memory for "
"Rx completion ring\n");
goto fail;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ctx.sf_busaddr = 0;
error = bus_dmamap_load(sc->sf_cdata.sf_rx_cring_tag,
sc->sf_cdata.sf_rx_cring_map, sc->sf_rdata.sf_rx_cring,
SF_RX_CLIST_SIZE, sf_dmamap_cb, &ctx, 0);
if (error != 0 || ctx.sf_busaddr == 0) {
device_printf(sc->sf_dev,
"failed to load DMA'able memory for Rx completion ring\n");
goto fail;
}
sc->sf_rdata.sf_rx_cring_paddr = ctx.sf_busaddr;
/*
* Tx desciptor ring and Tx completion ring should be addressed in
* the same 4GB space. The same rule applys to Rx ring and Rx
* completion ring. Unfortunately there is no way to specify this
* boundary restriction with bus_dma(9). So just try to allocate
* without the restriction and check the restriction was satisfied.
* If not, fall back to 32bit dma addressing mode which always
* guarantees the restriction.
*/
tx_ring_end = sc->sf_rdata.sf_tx_ring_paddr + SF_TX_DLIST_SIZE;
tx_cring_end = sc->sf_rdata.sf_tx_cring_paddr + SF_TX_CLIST_SIZE;
rx_ring_end = sc->sf_rdata.sf_rx_ring_paddr + SF_RX_DLIST_SIZE;
rx_cring_end = sc->sf_rdata.sf_rx_cring_paddr + SF_RX_CLIST_SIZE;
if ((SF_ADDR_HI(sc->sf_rdata.sf_tx_ring_paddr) !=
SF_ADDR_HI(tx_cring_end)) ||
(SF_ADDR_HI(sc->sf_rdata.sf_tx_cring_paddr) !=
SF_ADDR_HI(tx_ring_end)) ||
(SF_ADDR_HI(sc->sf_rdata.sf_rx_ring_paddr) !=
SF_ADDR_HI(rx_cring_end)) ||
(SF_ADDR_HI(sc->sf_rdata.sf_rx_cring_paddr) !=
SF_ADDR_HI(rx_ring_end))) {
device_printf(sc->sf_dev,
"switching to 32bit DMA mode\n");
sf_dma_free(sc);
/* Limit DMA address space to 32bit and try again. */
lowaddr = BUS_SPACE_MAXADDR_32BIT;
goto again;
}
/* Create DMA maps for Tx buffers. */
for (i = 0; i < SF_TX_DLIST_CNT; i++) {
txd = &sc->sf_cdata.sf_txdesc[i];
txd->tx_m = NULL;
txd->ndesc = 0;
txd->tx_dmamap = NULL;
error = bus_dmamap_create(sc->sf_cdata.sf_tx_tag, 0,
&txd->tx_dmamap);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to create Tx dmamap\n");
goto fail;
}
}
/* Create DMA maps for Rx buffers. */
if ((error = bus_dmamap_create(sc->sf_cdata.sf_rx_tag, 0,
&sc->sf_cdata.sf_rx_sparemap)) != 0) {
device_printf(sc->sf_dev,
"failed to create spare Rx dmamap\n");
goto fail;
}
for (i = 0; i < SF_RX_DLIST_CNT; i++) {
rxd = &sc->sf_cdata.sf_rxdesc[i];
rxd->rx_m = NULL;
rxd->rx_dmamap = NULL;
error = bus_dmamap_create(sc->sf_cdata.sf_rx_tag, 0,
&rxd->rx_dmamap);
if (error != 0) {
device_printf(sc->sf_dev,
"failed to create Rx dmamap\n");
goto fail;
}
}
fail:
return (error);
}
static void
sf_dma_free(struct sf_softc *sc)
{
struct sf_txdesc *txd;
struct sf_rxdesc *rxd;
int i;
/* Tx ring. */
if (sc->sf_cdata.sf_tx_ring_tag) {
if (sc->sf_cdata.sf_tx_ring_map)
bus_dmamap_unload(sc->sf_cdata.sf_tx_ring_tag,
sc->sf_cdata.sf_tx_ring_map);
if (sc->sf_cdata.sf_tx_ring_map &&
sc->sf_rdata.sf_tx_ring)
bus_dmamem_free(sc->sf_cdata.sf_tx_ring_tag,
sc->sf_rdata.sf_tx_ring,
sc->sf_cdata.sf_tx_ring_map);
sc->sf_rdata.sf_tx_ring = NULL;
sc->sf_cdata.sf_tx_ring_map = NULL;
bus_dma_tag_destroy(sc->sf_cdata.sf_tx_ring_tag);
sc->sf_cdata.sf_tx_ring_tag = NULL;
}
/* Tx completion ring. */
if (sc->sf_cdata.sf_tx_cring_tag) {
if (sc->sf_cdata.sf_tx_cring_map)
bus_dmamap_unload(sc->sf_cdata.sf_tx_cring_tag,
sc->sf_cdata.sf_tx_cring_map);
if (sc->sf_cdata.sf_tx_cring_map &&
sc->sf_rdata.sf_tx_cring)
bus_dmamem_free(sc->sf_cdata.sf_tx_cring_tag,
sc->sf_rdata.sf_tx_cring,
sc->sf_cdata.sf_tx_cring_map);
sc->sf_rdata.sf_tx_cring = NULL;
sc->sf_cdata.sf_tx_cring_map = NULL;
bus_dma_tag_destroy(sc->sf_cdata.sf_tx_cring_tag);
sc->sf_cdata.sf_tx_cring_tag = NULL;
}
/* Rx ring. */
if (sc->sf_cdata.sf_rx_ring_tag) {
if (sc->sf_cdata.sf_rx_ring_map)
bus_dmamap_unload(sc->sf_cdata.sf_rx_ring_tag,
sc->sf_cdata.sf_rx_ring_map);
if (sc->sf_cdata.sf_rx_ring_map &&
sc->sf_rdata.sf_rx_ring)
bus_dmamem_free(sc->sf_cdata.sf_rx_ring_tag,
sc->sf_rdata.sf_rx_ring,
sc->sf_cdata.sf_rx_ring_map);
sc->sf_rdata.sf_rx_ring = NULL;
sc->sf_cdata.sf_rx_ring_map = NULL;
bus_dma_tag_destroy(sc->sf_cdata.sf_rx_ring_tag);
sc->sf_cdata.sf_rx_ring_tag = NULL;
}
/* Rx completion ring. */
if (sc->sf_cdata.sf_rx_cring_tag) {
if (sc->sf_cdata.sf_rx_cring_map)
bus_dmamap_unload(sc->sf_cdata.sf_rx_cring_tag,
sc->sf_cdata.sf_rx_cring_map);
if (sc->sf_cdata.sf_rx_cring_map &&
sc->sf_rdata.sf_rx_cring)
bus_dmamem_free(sc->sf_cdata.sf_rx_cring_tag,
sc->sf_rdata.sf_rx_cring,
sc->sf_cdata.sf_rx_cring_map);
sc->sf_rdata.sf_rx_cring = NULL;
sc->sf_cdata.sf_rx_cring_map = NULL;
bus_dma_tag_destroy(sc->sf_cdata.sf_rx_cring_tag);
sc->sf_cdata.sf_rx_cring_tag = NULL;
}
/* Tx buffers. */
if (sc->sf_cdata.sf_tx_tag) {
for (i = 0; i < SF_TX_DLIST_CNT; i++) {
txd = &sc->sf_cdata.sf_txdesc[i];
if (txd->tx_dmamap) {
bus_dmamap_destroy(sc->sf_cdata.sf_tx_tag,
txd->tx_dmamap);
txd->tx_dmamap = NULL;
}
}
bus_dma_tag_destroy(sc->sf_cdata.sf_tx_tag);
sc->sf_cdata.sf_tx_tag = NULL;
}
/* Rx buffers. */
if (sc->sf_cdata.sf_rx_tag) {
for (i = 0; i < SF_RX_DLIST_CNT; i++) {
rxd = &sc->sf_cdata.sf_rxdesc[i];
if (rxd->rx_dmamap) {
bus_dmamap_destroy(sc->sf_cdata.sf_rx_tag,
rxd->rx_dmamap);
rxd->rx_dmamap = NULL;
}
}
if (sc->sf_cdata.sf_rx_sparemap) {
bus_dmamap_destroy(sc->sf_cdata.sf_rx_tag,
sc->sf_cdata.sf_rx_sparemap);
sc->sf_cdata.sf_rx_sparemap = 0;
}
bus_dma_tag_destroy(sc->sf_cdata.sf_rx_tag);
sc->sf_cdata.sf_rx_tag = NULL;
}
if (sc->sf_cdata.sf_parent_tag) {
bus_dma_tag_destroy(sc->sf_cdata.sf_parent_tag);
sc->sf_cdata.sf_parent_tag = NULL;
}
}
static int
sf_init_rx_ring(struct sf_softc *sc)
{
struct sf_ring_data *rd;
int i;
sc->sf_cdata.sf_rxc_cons = 0;
rd = &sc->sf_rdata;
bzero(rd->sf_rx_ring, SF_RX_DLIST_SIZE);
bzero(rd->sf_rx_cring, SF_RX_CLIST_SIZE);
for (i = 0; i < SF_RX_DLIST_CNT; i++) {
if (sf_newbuf(sc, i) != 0)
return (ENOBUFS);
}
bus_dmamap_sync(sc->sf_cdata.sf_rx_cring_tag,
sc->sf_cdata.sf_rx_cring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
bus_dmamap_sync(sc->sf_cdata.sf_rx_ring_tag,
sc->sf_cdata.sf_rx_ring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
return (0);
}
static void
sf_init_tx_ring(struct sf_softc *sc)
{
struct sf_ring_data *rd;
int i;
sc->sf_cdata.sf_tx_prod = 0;
sc->sf_cdata.sf_tx_cnt = 0;
sc->sf_cdata.sf_txc_cons = 0;
rd = &sc->sf_rdata;
bzero(rd->sf_tx_ring, SF_TX_DLIST_SIZE);
bzero(rd->sf_tx_cring, SF_TX_CLIST_SIZE);
for (i = 0; i < SF_TX_DLIST_CNT; i++) {
rd->sf_tx_ring[i].sf_tx_ctrl = htole32(SF_TX_DESC_ID);
sc->sf_cdata.sf_txdesc[i].tx_m = NULL;
sc->sf_cdata.sf_txdesc[i].ndesc = 0;
}
rd->sf_tx_ring[i].sf_tx_ctrl |= htole32(SF_TX_DESC_END);
bus_dmamap_sync(sc->sf_cdata.sf_tx_ring_tag,
sc->sf_cdata.sf_tx_ring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
bus_dmamap_sync(sc->sf_cdata.sf_tx_cring_tag,
sc->sf_cdata.sf_tx_cring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
}
/*
* Initialize an RX descriptor and attach an MBUF cluster.
*/
static int
sf_newbuf(struct sf_softc *sc, int idx)
{
struct sf_rx_rdesc *desc;
struct sf_rxdesc *rxd;
struct mbuf *m;
bus_dma_segment_t segs[1];
bus_dmamap_t map;
int nsegs;
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (m == NULL)
return (ENOBUFS);
m->m_len = m->m_pkthdr.len = MCLBYTES;
m_adj(m, sizeof(uint32_t));
if (bus_dmamap_load_mbuf_sg(sc->sf_cdata.sf_rx_tag,
sc->sf_cdata.sf_rx_sparemap, m, segs, &nsegs, 0) != 0) {
m_freem(m);
return (ENOBUFS);
}
KASSERT(nsegs == 1, ("%s: %d segments returned!", __func__, nsegs));
rxd = &sc->sf_cdata.sf_rxdesc[idx];
if (rxd->rx_m != NULL) {
bus_dmamap_sync(sc->sf_cdata.sf_rx_tag, rxd->rx_dmamap,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->sf_cdata.sf_rx_tag, rxd->rx_dmamap);
}
map = rxd->rx_dmamap;
rxd->rx_dmamap = sc->sf_cdata.sf_rx_sparemap;
sc->sf_cdata.sf_rx_sparemap = map;
bus_dmamap_sync(sc->sf_cdata.sf_rx_tag, rxd->rx_dmamap,
BUS_DMASYNC_PREREAD);
rxd->rx_m = m;
desc = &sc->sf_rdata.sf_rx_ring[idx];
desc->sf_addr = htole64(segs[0].ds_addr);
return (0);
}
#ifndef __NO_STRICT_ALIGNMENT
static __inline void
sf_fixup_rx(struct mbuf *m)
{
int i;
uint16_t *src, *dst;
src = mtod(m, uint16_t *);
dst = src - 1;
for (i = 0; i < (m->m_len / sizeof(uint16_t) + 1); i++)
*dst++ = *src++;
m->m_data -= ETHER_ALIGN;
}
#endif
/*
* The starfire is programmed to use 'normal' mode for packet reception,
* which means we use the consumer/producer model for both the buffer
* descriptor queue and the completion descriptor queue. The only problem
* with this is that it involves a lot of register accesses: we have to
* read the RX completion consumer and producer indexes and the RX buffer
* producer index, plus the RX completion consumer and RX buffer producer
* indexes have to be updated. It would have been easier if Adaptec had
* put each index in a separate register, especially given that the damn
* NIC has a 512K register space.
*
* In spite of all the lovely features that Adaptec crammed into the 6915,
* it is marred by one truly stupid design flaw, which is that receive
* buffer addresses must be aligned on a longword boundary. This forces
* the packet payload to be unaligned, which is suboptimal on the x86 and
* completely unuseable on the Alpha. Our only recourse is to copy received
* packets into properly aligned buffers before handing them off.
*/
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_rxeof(struct sf_softc *sc)
{
struct mbuf *m;
struct ifnet *ifp;
struct sf_rxdesc *rxd;
struct sf_rx_rcdesc *cur_cmp;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int cons, eidx, prog, rx_npkts;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint32_t status, status2;
SF_LOCK_ASSERT(sc);
ifp = sc->sf_ifp;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = 0;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
bus_dmamap_sync(sc->sf_cdata.sf_rx_ring_tag,
sc->sf_cdata.sf_rx_ring_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
bus_dmamap_sync(sc->sf_cdata.sf_rx_cring_tag,
sc->sf_cdata.sf_rx_cring_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
/*
* To reduce register access, directly read Receive completion
* queue entry.
*/
eidx = 0;
prog = 0;
Add check for IFF_DRV_RUNNING flag after serving an interrupt and don't give RX path more priority than TX path. Also remove infinite loop in interrupt handler and limit number of iteration to 32. This change addresses system load fluctuations under high network load.
2012-02-23 08:22:44 +00:00
for (cons = sc->sf_cdata.sf_rxc_cons;
(ifp->if_drv_flags & IFF_DRV_RUNNING) != 0;
SF_INC(cons, SF_RX_CLIST_CNT)) {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
cur_cmp = &sc->sf_rdata.sf_rx_cring[cons];
status = le32toh(cur_cmp->sf_rx_status1);
if (status == 0)
break;
#ifdef DEVICE_POLLING
if ((ifp->if_capenable & IFCAP_POLLING) != 0) {
if (sc->rxcycles <= 0)
break;
sc->rxcycles--;
}
#endif
prog++;
eidx = (status & SF_RX_CMPDESC_EIDX) >> 16;
rxd = &sc->sf_cdata.sf_rxdesc[eidx];
m = rxd->rx_m;
/*
* Note, if_ipackets and if_ierrors counters
* are handled in sf_stats_update().
*/
if ((status & SF_RXSTAT1_OK) == 0) {
cur_cmp->sf_rx_status1 = 0;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
continue;
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sf_newbuf(sc, eidx) != 0) {
ifp->if_iqdrops++;
cur_cmp->sf_rx_status1 = 0;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
continue;
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* AIC-6915 supports TCP/UDP checksum offload. */
if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) {
status2 = le32toh(cur_cmp->sf_rx_status2);
/*
* Sometimes AIC-6915 generates an interrupt to
* warn RxGFP stall with bad checksum bit set
* in status word. I'm not sure what conditioan
* triggers it but recevied packet's checksum
* was correct even though AIC-6915 does not
* agree on this. This may be an indication of
* firmware bug. To fix the issue, do not rely
* on bad checksum bit in status word and let
* upper layer verify integrity of received
* frame.
* Another nice feature of AIC-6915 is hardware
* assistance of checksum calculation by
* providing partial checksum value for received
* frame. The partial checksum value can be used
* to accelerate checksum computation for
* fragmented TCP/UDP packets. Upper network
* stack already takes advantage of the partial
* checksum value in IP reassembly stage. But
* I'm not sure the correctness of the partial
* hardware checksum assistance as frequent
* RxGFP stalls are seen on non-fragmented
* frames. Due to the nature of the complexity
* of checksum computation code in firmware it's
* possible to see another bug in RxGFP so
* ignore checksum assistance for fragmented
* frames. This can be changed in future.
*/
if ((status2 & SF_RXSTAT2_FRAG) == 0) {
if ((status2 & (SF_RXSTAT2_TCP |
SF_RXSTAT2_UDP)) != 0) {
if ((status2 & SF_RXSTAT2_CSUM_OK)) {
m->m_pkthdr.csum_flags =
CSUM_DATA_VALID |
CSUM_PSEUDO_HDR;
m->m_pkthdr.csum_data = 0xffff;
}
}
}
#ifdef SF_PARTIAL_CSUM_SUPPORT
else if ((status2 & SF_RXSTAT2_FRAG) != 0) {
if ((status2 & (SF_RXSTAT2_TCP |
SF_RXSTAT2_UDP)) != 0) {
if ((status2 & SF_RXSTAT2_PCSUM_OK)) {
m->m_pkthdr.csum_flags =
CSUM_DATA_VALID;
m->m_pkthdr.csum_data =
(status &
SF_RX_CMPDESC_CSUM2);
}
}
}
#endif
}
m->m_pkthdr.len = m->m_len = status & SF_RX_CMPDESC_LEN;
#ifndef __NO_STRICT_ALIGNMENT
sf_fixup_rx(m);
#endif
m->m_pkthdr.rcvif = ifp;
Drop the driver lock around calls to if_input to avoid a LOR when the packets are immediately returned for sending (e.g. when bridging or packet forwarding). There are more efficient ways to do this but for now use the least intrusive approach. Reviewed by: imp, rwatson
2003-11-14 19:00:32 +00:00
SF_UNLOCK(sc);
o track either_ifattach/ether_ifdetach API changes o use if_input for input packet processing o don't strip the Ethernet header for input packets o use BPF_* macros bpf tapping o call ether_ioctl to handle default ioctl case o track vlan changes Reviewed by: many Approved by: re
2002-11-14 23:49:09 +00:00
(*ifp->if_input)(ifp, m);
Drop the driver lock around calls to if_input to avoid a LOR when the packets are immediately returned for sending (e.g. when bridging or packet forwarding). There are more efficient ways to do this but for now use the least intrusive approach. Reviewed by: imp, rwatson
2003-11-14 19:00:32 +00:00
SF_LOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts++;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Clear completion status. */
cur_cmp->sf_rx_status1 = 0;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (prog > 0) {
sc->sf_cdata.sf_rxc_cons = cons;
bus_dmamap_sync(sc->sf_cdata.sf_rx_ring_tag,
sc->sf_cdata.sf_rx_ring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
bus_dmamap_sync(sc->sf_cdata.sf_rx_cring_tag,
sc->sf_cdata.sf_rx_cring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
/* Update Rx completion Q1 consumer index. */
csr_write_4(sc, SF_CQ_CONSIDX,
(csr_read_4(sc, SF_CQ_CONSIDX) & ~SF_CQ_CONSIDX_RXQ1) |
(cons & SF_CQ_CONSIDX_RXQ1));
/* Update Rx descriptor Q1 ptr. */
csr_write_4(sc, SF_RXDQ_PTR_Q1,
(csr_read_4(sc, SF_RXDQ_PTR_Q1) & ~SF_RXDQ_PRODIDX) |
(eidx & SF_RXDQ_PRODIDX));
}
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
/*
* Read the transmit status from the completion queue and release
* mbufs. Note that the buffer descriptor index in the completion
* descriptor is an offset from the start of the transmit buffer
* descriptor list in bytes. This is important because the manual
* gives the impression that it should match the producer/consumer
* index, which is the offset in 8 byte blocks.
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_txeof(struct sf_softc *sc)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_txdesc *txd;
struct sf_tx_rcdesc *cur_cmp;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
struct ifnet *ifp;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint32_t status;
int cons, idx, prod;
SF_LOCK_ASSERT(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->sf_ifp;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
bus_dmamap_sync(sc->sf_cdata.sf_tx_cring_tag,
sc->sf_cdata.sf_tx_cring_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
cons = sc->sf_cdata.sf_txc_cons;
prod = (csr_read_4(sc, SF_CQ_PRODIDX) & SF_TXDQ_PRODIDX_HIPRIO) >> 16;
if (prod == cons)
return;
for (; cons != prod; SF_INC(cons, SF_TX_CLIST_CNT)) {
cur_cmp = &sc->sf_rdata.sf_tx_cring[cons];
status = le32toh(cur_cmp->sf_tx_status1);
if (status == 0)
break;
switch (status & SF_TX_CMPDESC_TYPE) {
case SF_TXCMPTYPE_TX:
/* Tx complete entry. */
break;
case SF_TXCMPTYPE_DMA:
/* DMA complete entry. */
idx = status & SF_TX_CMPDESC_IDX;
idx = idx / sizeof(struct sf_tx_rdesc);
/*
* We don't need to check Tx status here.
* SF_ISR_TX_LOFIFO intr would handle this.
* Note, if_opackets, if_collisions and if_oerrors
* counters are handled in sf_stats_update().
*/
txd = &sc->sf_cdata.sf_txdesc[idx];
if (txd->tx_m != NULL) {
bus_dmamap_sync(sc->sf_cdata.sf_tx_tag,
txd->tx_dmamap,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(sc->sf_cdata.sf_tx_tag,
txd->tx_dmamap);
m_freem(txd->tx_m);
txd->tx_m = NULL;
}
sc->sf_cdata.sf_tx_cnt -= txd->ndesc;
KASSERT(sc->sf_cdata.sf_tx_cnt >= 0,
("%s: Active Tx desc counter was garbled\n",
__func__));
txd->ndesc = 0;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Still more changes to try to prevent TX lockups. Will wait for one more night of testing before merging to -stable. Also added to code to detect TX underruns and automatically increase the TX threshold to avoid them. Carefully placed diagnostig printf() about this under #ifdef DIAGNOSTIC to avoid getting any panicky e-mails from confused users, like I always do with the xl and dc drivers.
2001-08-16 00:32:20 +00:00
break;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
default:
/* It should not happen. */
device_printf(sc->sf_dev,
"unknown Tx completion type : 0x%08x : %d : %d\n",
status, cons, prod);
break;
}
cur_cmp->sf_tx_status1 = 0;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc->sf_cdata.sf_txc_cons = cons;
bus_dmamap_sync(sc->sf_cdata.sf_tx_cring_tag,
sc->sf_cdata.sf_tx_cring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_cdata.sf_tx_cnt == 0)
sc->sf_watchdog_timer = 0;
/* Update Tx completion consumer index. */
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_CQ_CONSIDX,
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
(csr_read_4(sc, SF_CQ_CONSIDX) & 0xffff) |
((cons << 16) & 0xffff0000));
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_txthresh_adjust(struct sf_softc *sc)
Still more changes to try to prevent TX lockups. Will wait for one more night of testing before merging to -stable. Also added to code to detect TX underruns and automatically increase the TX threshold to avoid them. Carefully placed diagnostig printf() about this under #ifdef DIAGNOSTIC to avoid getting any panicky e-mails from confused users, like I always do with the xl and dc drivers.
2001-08-16 00:32:20 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint32_t txfctl;
device_printf(sc->sf_dev, "Tx underrun -- ");
if (sc->sf_txthresh < SF_MAX_TX_THRESHOLD) {
txfctl = csr_read_4(sc, SF_TX_FRAMCTL);
/* Increase Tx threshold 256 bytes. */
sc->sf_txthresh += 16;
if (sc->sf_txthresh > SF_MAX_TX_THRESHOLD)
sc->sf_txthresh = SF_MAX_TX_THRESHOLD;
Still more changes to try to prevent TX lockups. Will wait for one more night of testing before merging to -stable. Also added to code to detect TX underruns and automatically increase the TX threshold to avoid them. Carefully placed diagnostig printf() about this under #ifdef DIAGNOSTIC to avoid getting any panicky e-mails from confused users, like I always do with the xl and dc drivers.
2001-08-16 00:32:20 +00:00
txfctl &= ~SF_TXFRMCTL_TXTHRESH;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
txfctl |= sc->sf_txthresh;
printf("increasing Tx threshold to %d bytes\n",
sc->sf_txthresh * SF_TX_THRESHOLD_UNIT);
Still more changes to try to prevent TX lockups. Will wait for one more night of testing before merging to -stable. Also added to code to detect TX underruns and automatically increase the TX threshold to avoid them. Carefully placed diagnostig printf() about this under #ifdef DIAGNOSTIC to avoid getting any panicky e-mails from confused users, like I always do with the xl and dc drivers.
2001-08-16 00:32:20 +00:00
csr_write_4(sc, SF_TX_FRAMCTL, txfctl);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
} else
printf("\n");
Still more changes to try to prevent TX lockups. Will wait for one more night of testing before merging to -stable. Also added to code to detect TX underruns and automatically increase the TX threshold to avoid them. Carefully placed diagnostig printf() about this under #ifdef DIAGNOSTIC to avoid getting any panicky e-mails from confused users, like I always do with the xl and dc drivers.
2001-08-16 00:32:20 +00:00
}
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
#ifdef DEVICE_POLLING
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
sf_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_softc *sc;
uint32_t status;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int rx_npkts;
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc = ifp->if_softc;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = 0;
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
SF_LOCK(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
SF_UNLOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
}
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
sc->rxcycles = count;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = sf_rxeof(sc);
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
sf_txeof(sc);
Add altq support. Patch by mlaier. Approved by: mlaier MFC after: 2 weeks
2004-11-12 18:12:04 +00:00
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
sf_start_locked(ifp);
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
if (cmd == POLL_AND_CHECK_STATUS) {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Reading the ISR register clears all interrrupts. */
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
status = csr_read_4(sc, SF_ISR);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((status & SF_ISR_ABNORMALINTR) != 0) {
if ((status & SF_ISR_STATSOFLOW) != 0)
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
sf_stats_update(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
else if ((status & SF_ISR_TX_LOFIFO) != 0)
sf_txthresh_adjust(sc);
else if ((status & SF_ISR_DMAERR) != 0) {
device_printf(sc->sf_dev,
"DMA error, resetting\n");
Remove unnecessary controller reinitialization. StarFire controller does not require controller reinitialization to program perfect filters. While here, make driver immediately exit from interrupt/polling handler if driver reinitialized controller. PR: kern/87506
2010-09-21 17:42:19 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
sf_init_locked(sc);
Remove unnecessary controller reinitialization. StarFire controller does not require controller reinitialization to program perfect filters. While here, make driver immediately exit from interrupt/polling handler if driver reinitialized controller. PR: kern/87506
2010-09-21 17:42:19 +00:00
SF_UNLOCK(sc);
return (rx_npkts);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
} else if ((status & SF_ISR_NO_TX_CSUM) != 0) {
Fix build.
2008-01-21 09:51:28 +00:00
sc->sf_statistics.sf_tx_gfp_stall++;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#ifdef SF_GFP_DEBUG
device_printf(sc->sf_dev,
"TxGFP is not responding!\n");
#endif
} else if ((status & SF_ISR_RXGFP_NORESP) != 0) {
Fix build.
2008-01-21 09:51:28 +00:00
sc->sf_statistics.sf_rx_gfp_stall++;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#ifdef SF_GFP_DEBUG
device_printf(sc->sf_dev,
"RxGFP is not responding!\n");
#endif
}
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
}
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
SF_UNLOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
}
#endif /* DEVICE_POLLING */
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_intr(void *arg)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct sf_softc *sc;
struct ifnet *ifp;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint32_t status;
Add check for IFF_DRV_RUNNING flag after serving an interrupt and don't give RX path more priority than TX path. Also remove infinite loop in interrupt handler and limit number of iteration to 32. This change addresses system load fluctuations under high network load.
2012-02-23 08:22:44 +00:00
int cnt;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc = (struct sf_softc *)arg;
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
SF_LOCK(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_suspended != 0)
goto done_locked;
/* Reading the ISR register clears all interrrupts. */
status = csr_read_4(sc, SF_ISR);
if (status == 0 || status == 0xffffffff ||
(status & SF_ISR_PCIINT_ASSERTED) == 0)
goto done_locked;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ifp = sc->sf_ifp;
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
#ifdef DEVICE_POLLING
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((ifp->if_capenable & IFCAP_POLLING) != 0)
goto done_locked;
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#endif
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* Disable interrupts. */
csr_write_4(sc, SF_IMR, 0x00000000);
Add check for IFF_DRV_RUNNING flag after serving an interrupt and don't give RX path more priority than TX path. Also remove infinite loop in interrupt handler and limit number of iteration to 32. This change addresses system load fluctuations under high network load.
2012-02-23 08:22:44 +00:00
for (cnt = 32; (status & SF_INTRS) != 0;) {
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
break;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((status & SF_ISR_RXDQ1_DMADONE) != 0)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_rxeof(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((status & (SF_ISR_TX_TXDONE | SF_ISR_TX_DMADONE |
SF_ISR_TX_QUEUEDONE)) != 0)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_txeof(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((status & SF_ISR_ABNORMALINTR) != 0) {
if ((status & SF_ISR_STATSOFLOW) != 0)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_stats_update(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
else if ((status & SF_ISR_TX_LOFIFO) != 0)
sf_txthresh_adjust(sc);
else if ((status & SF_ISR_DMAERR) != 0) {
device_printf(sc->sf_dev,
"DMA error, resetting\n");
Remove unnecessary controller reinitialization. StarFire controller does not require controller reinitialization to program perfect filters. While here, make driver immediately exit from interrupt/polling handler if driver reinitialized controller. PR: kern/87506
2010-09-21 17:42:19 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
sf_init_locked(sc);
Remove unnecessary controller reinitialization. StarFire controller does not require controller reinitialization to program perfect filters. While here, make driver immediately exit from interrupt/polling handler if driver reinitialized controller. PR: kern/87506
2010-09-21 17:42:19 +00:00
SF_UNLOCK(sc);
return;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
} else if ((status & SF_ISR_NO_TX_CSUM) != 0) {
sc->sf_statistics.sf_tx_gfp_stall++;
#ifdef SF_GFP_DEBUG
device_printf(sc->sf_dev,
"TxGFP is not responding!\n");
#endif
}
else if ((status & SF_ISR_RXGFP_NORESP) != 0) {
sc->sf_statistics.sf_rx_gfp_stall++;
#ifdef SF_GFP_DEBUG
device_printf(sc->sf_dev,
"RxGFP is not responding!\n");
#endif
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Add check for IFF_DRV_RUNNING flag after serving an interrupt and don't give RX path more priority than TX path. Also remove infinite loop in interrupt handler and limit number of iteration to 32. This change addresses system load fluctuations under high network load.
2012-02-23 08:22:44 +00:00
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
sf_start_locked(ifp);
if (--cnt <= 0)
break;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Reading the ISR register clears all interrrupts. */
status = csr_read_4(sc, SF_ISR);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Add check for IFF_DRV_RUNNING flag after serving an interrupt and don't give RX path more priority than TX path. Also remove infinite loop in interrupt handler and limit number of iteration to 32. This change addresses system load fluctuations under high network load.
2012-02-23 08:22:44 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) {
/* Re-enable interrupts. */
csr_write_4(sc, SF_IMR, SF_INTRS);
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
done_locked:
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
SF_UNLOCK(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_download_fw(struct sf_softc *sc)
{
uint32_t gfpinst;
int i, ndx;
uint8_t *p;
/*
* A FP instruction is composed of 48bits so we have to
* write it with two parts.
*/
p = txfwdata;
ndx = 0;
for (i = 0; i < sizeof(txfwdata) / SF_GFP_INST_BYTES; i++) {
gfpinst = p[2] << 24 | p[3] << 16 | p[4] << 8 | p[5];
csr_write_4(sc, SF_TXGFP_MEM_BASE + ndx * 4, gfpinst);
gfpinst = p[0] << 8 | p[1];
csr_write_4(sc, SF_TXGFP_MEM_BASE + (ndx + 1) * 4, gfpinst);
p += SF_GFP_INST_BYTES;
ndx += 2;
}
if (bootverbose)
device_printf(sc->sf_dev, "%d Tx instructions downloaded\n", i);
p = rxfwdata;
ndx = 0;
for (i = 0; i < sizeof(rxfwdata) / SF_GFP_INST_BYTES; i++) {
gfpinst = p[2] << 24 | p[3] << 16 | p[4] << 8 | p[5];
csr_write_4(sc, SF_RXGFP_MEM_BASE + (ndx * 4), gfpinst);
gfpinst = p[0] << 8 | p[1];
csr_write_4(sc, SF_RXGFP_MEM_BASE + (ndx + 1) * 4, gfpinst);
p += SF_GFP_INST_BYTES;
ndx += 2;
}
if (bootverbose)
device_printf(sc->sf_dev, "%d Rx instructions downloaded\n", i);
}
static void
sf_init(void *xsc)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct sf_softc *sc;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc = (struct sf_softc *)xsc;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK(sc);
sf_init_locked(sc);
SF_UNLOCK(sc);
}
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_init_locked(struct sf_softc *sc)
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
{
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
struct ifnet *ifp;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint8_t eaddr[ETHER_ADDR_LEN];
bus_addr_t addr;
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
int i;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->sf_ifp;
Remove unnecessary controller reinitialization. StarFire controller does not require controller reinitialization to program perfect filters. While here, make driver immediately exit from interrupt/polling handler if driver reinitialized controller. PR: kern/87506
2010-09-21 17:42:19 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
return;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_stop(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Reset the hardware to a known state. */
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_reset(sc);
/* Init all the receive filter registers */
for (i = SF_RXFILT_PERFECT_BASE;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
i < (SF_RXFILT_HASH_MAX + 1); i += sizeof(uint32_t))
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, i, 0);
/* Empty stats counter registers. */
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
for (i = SF_STATS_BASE; i < (SF_STATS_END + 1); i += sizeof(uint32_t))
csr_write_4(sc, i, 0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Init our MAC address. */
bcopy(IF_LLADDR(sc->sf_ifp), eaddr, sizeof(eaddr));
csr_write_4(sc, SF_PAR0,
eaddr[2] << 24 | eaddr[3] << 16 | eaddr[4] << 8 | eaddr[5]);
csr_write_4(sc, SF_PAR1, eaddr[0] << 8 | eaddr[1]);
sf_setperf(sc, 0, eaddr);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
if (sf_init_rx_ring(sc) == ENOBUFS) {
Consistently use if_printf() only in interface methods: if_start, if_ioctl, if_watchdog, etc, or in functions that are used by these methods only. In all other cases use device_printf(). This also fixes several panics, when if_printf() is called before softc->ifp was initialized. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 10:40:54 +00:00
device_printf(sc->sf_dev,
Use if_printf() and device_printf() and axe sf_unit from the softc as a result.
2005-08-10 20:24:39 +00:00
"initialization failed: no memory for rx buffers\n");
If there are not enough RX buffers, release partially allocated RX buffers.
2012-02-23 05:14:11 +00:00
sf_stop(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
return;
}
sf_init_tx_ring(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* 16 perfect address filtering.
* Hash only multicast destination address, Accept matching
* frames regardless of VLAN ID.
*/
csr_write_4(sc, SF_RXFILT, SF_PERFMODE_NORMAL | SF_HASHMODE_ANYVLAN);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Close PR# 19904: remember to call sf_setmulti() when initializing the interface so the multicast filter will be programmed correctly.
2000-07-14 19:11:02 +00:00
/*
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
* Set Rx filter.
Close PR# 19904: remember to call sf_setmulti() when initializing the interface so the multicast filter will be programmed correctly.
2000-07-14 19:11:02 +00:00
*/
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_rxfilter(sc);
Close PR# 19904: remember to call sf_setmulti() when initializing the interface so the multicast filter will be programmed correctly.
2000-07-14 19:11:02 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Init the completion queue indexes. */
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_CQ_CONSIDX, 0);
csr_write_4(sc, SF_CQ_PRODIDX, 0);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Init the RX completion queue. */
addr = sc->sf_rdata.sf_rx_cring_paddr;
csr_write_4(sc, SF_CQ_ADDR_HI, SF_ADDR_HI(addr));
csr_write_4(sc, SF_RXCQ_CTL_1, SF_ADDR_LO(addr) & SF_RXCQ_ADDR);
if (SF_ADDR_HI(addr) != 0)
SF_SETBIT(sc, SF_RXCQ_CTL_1, SF_RXCQ_USE_64BIT);
/* Set RX completion queue type 2. */
SF_SETBIT(sc, SF_RXCQ_CTL_1, SF_RXCQTYPE_2);
csr_write_4(sc, SF_RXCQ_CTL_2, 0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* Init RX DMA control.
* default RxHighPriority Threshold,
* default RxBurstSize, 128bytes.
*/
SF_SETBIT(sc, SF_RXDMA_CTL,
SF_RXDMA_REPORTBADPKTS |
(SF_RXDMA_HIGHPRIO_THRESH << 8) |
SF_RXDMA_BURST);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* Init the RX buffer descriptor queue. */
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
addr = sc->sf_rdata.sf_rx_ring_paddr;
csr_write_4(sc, SF_RXDQ_ADDR_HI, SF_ADDR_HI(addr));
csr_write_4(sc, SF_RXDQ_ADDR_Q1, SF_ADDR_LO(addr));
/* Set RX queue buffer length. */
csr_write_4(sc, SF_RXDQ_CTL_1,
((MCLBYTES - sizeof(uint32_t)) << 16) |
SF_RXDQCTL_64BITBADDR | SF_RXDQCTL_VARIABLE);
if (SF_ADDR_HI(addr) != 0)
SF_SETBIT(sc, SF_RXDQ_CTL_1, SF_RXDQCTL_64BITDADDR);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_RXDQ_PTR_Q1, SF_RX_DLIST_CNT - 1);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
csr_write_4(sc, SF_RXDQ_CTL_2, 0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* Init the TX completion queue */
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
addr = sc->sf_rdata.sf_tx_cring_paddr;
csr_write_4(sc, SF_TXCQ_CTL, SF_ADDR_LO(addr) & SF_TXCQ_ADDR);
if (SF_ADDR_HI(addr) != 0)
SF_SETBIT(sc, SF_TXCQ_CTL, SF_TXCQ_USE_64BIT);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* Init the TX buffer descriptor queue. */
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
addr = sc->sf_rdata.sf_tx_ring_paddr;
csr_write_4(sc, SF_TXDQ_ADDR_HI, SF_ADDR_HI(addr));
csr_write_4(sc, SF_TXDQ_ADDR_HIPRIO, 0);
csr_write_4(sc, SF_TXDQ_ADDR_LOPRIO, SF_ADDR_LO(addr));
csr_write_4(sc, SF_TX_FRAMCTL,
SF_TXFRMCTL_CPLAFTERTX | sc->sf_txthresh);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_TXDQ_CTL,
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
SF_TXDMA_HIPRIO_THRESH << 24 |
SF_TXSKIPLEN_0BYTES << 16 |
SF_TXDDMA_BURST << 8 |
SF_TXBUFDESC_TYPE2 | SF_TXMINSPACE_UNLIMIT);
if (SF_ADDR_HI(addr) != 0)
SF_SETBIT(sc, SF_TXDQ_CTL, SF_TXDQCTL_64BITADDR);
/* Set VLAN Type register. */
csr_write_4(sc, SF_VLANTYPE, ETHERTYPE_VLAN);
/* Set TxPause Timer. */
csr_write_4(sc, SF_TXPAUSETIMER, 0xffff);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* Enable autopadding of short TX frames. */
SF_SETBIT(sc, SF_MACCFG_1, SF_MACCFG1_AUTOPAD);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
SF_SETBIT(sc, SF_MACCFG_2, SF_MACCFG2_AUTOVLANPAD);
/* Make sure to reset MAC to take changes effect. */
SF_SETBIT(sc, SF_MACCFG_1, SF_MACCFG1_SOFTRESET);
DELAY(1000);
SF_CLRBIT(sc, SF_MACCFG_1, SF_MACCFG1_SOFTRESET);
/* Enable PCI bus master. */
SF_SETBIT(sc, SF_PCI_DEVCFG, SF_PCIDEVCFG_PCIMEN);
/* Load StarFire firmware. */
sf_download_fw(sc);
/* Intialize interrupt moderation. */
csr_write_4(sc, SF_TIMER_CTL, SF_TIMER_IMASK_MODE | SF_TIMER_TIMES_TEN |
(sc->sf_int_mod & SF_TIMER_IMASK_INTERVAL));
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
#ifdef DEVICE_POLLING
/* Disable interrupts if we are polling. */
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((ifp->if_capenable & IFCAP_POLLING) != 0)
Add device polling support Original patch by me, improvements by ru Happy birthday to: BSDforen.de! Approved by: ru MFC after: 2 weeks
2004-11-10 23:04:39 +00:00
csr_write_4(sc, SF_IMR, 0x00000000);
else
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#endif
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/* Enable interrupts. */
csr_write_4(sc, SF_IMR, SF_INTRS);
SF_SETBIT(sc, SF_PCI_DEVCFG, SF_PCIDEVCFG_INTR_ENB);
/* Enable the RX and TX engines. */
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
csr_write_4(sc, SF_GEN_ETH_CTL,
SF_ETHCTL_RX_ENB | SF_ETHCTL_RXDMA_ENB |
SF_ETHCTL_TX_ENB | SF_ETHCTL_TXDMA_ENB);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((ifp->if_capenable & IFCAP_TXCSUM) != 0)
SF_SETBIT(sc, SF_GEN_ETH_CTL, SF_ETHCTL_TXGFP_ENB);
else
SF_CLRBIT(sc, SF_GEN_ETH_CTL, SF_ETHCTL_TXGFP_ENB);
if ((ifp->if_capenable & IFCAP_RXCSUM) != 0)
SF_SETBIT(sc, SF_GEN_ETH_CTL, SF_ETHCTL_RXGFP_ENB);
else
SF_CLRBIT(sc, SF_GEN_ETH_CTL, SF_ETHCTL_RXGFP_ENB);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Introduce sf_ifmedia_upd_locked() and have driver reset PHY before switching to selected media. While here, set if_drv_flags before switching to selected media.
2012-02-23 05:23:21 +00:00
sc->sf_link = 0;
sf_ifmedia_upd_locked(ifp);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
callout_reset(&sc->sf_co, hz, sf_tick, sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_encap(struct sf_softc *sc, struct mbuf **m_head)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_txdesc *txd;
struct sf_tx_rdesc *desc;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
struct mbuf *m;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
bus_dmamap_t map;
bus_dma_segment_t txsegs[SF_MAXTXSEGS];
int error, i, nsegs, prod, si;
int avail, nskip;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
SF_LOCK_ASSERT(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
m = *m_head;
prod = sc->sf_cdata.sf_tx_prod;
txd = &sc->sf_cdata.sf_txdesc[prod];
map = txd->tx_dmamap;
error = bus_dmamap_load_mbuf_sg(sc->sf_cdata.sf_tx_tag, map,
*m_head, txsegs, &nsegs, BUS_DMA_NOWAIT);
if (error == EFBIG) {
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_collapse(*m_head, M_NOWAIT, SF_MAXTXSEGS);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (m == NULL) {
m_freem(*m_head);
*m_head = NULL;
return (ENOBUFS);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
*m_head = m;
error = bus_dmamap_load_mbuf_sg(sc->sf_cdata.sf_tx_tag,
map, *m_head, txsegs, &nsegs, BUS_DMA_NOWAIT);
if (error != 0) {
m_freem(*m_head);
*m_head = NULL;
return (error);
}
} else if (error != 0)
return (error);
if (nsegs == 0) {
m_freem(*m_head);
*m_head = NULL;
return (EIO);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Check number of available descriptors. */
avail = (SF_TX_DLIST_CNT - 1) - sc->sf_cdata.sf_tx_cnt;
if (avail < nsegs) {
bus_dmamap_unload(sc->sf_cdata.sf_tx_tag, map);
return (ENOBUFS);
}
nskip = 0;
if (prod + nsegs >= SF_TX_DLIST_CNT) {
nskip = SF_TX_DLIST_CNT - prod - 1;
if (avail < nsegs + nskip) {
bus_dmamap_unload(sc->sf_cdata.sf_tx_tag, map);
return (ENOBUFS);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
bus_dmamap_sync(sc->sf_cdata.sf_tx_tag, map, BUS_DMASYNC_PREWRITE);
si = prod;
for (i = 0; i < nsegs; i++) {
desc = &sc->sf_rdata.sf_tx_ring[prod];
desc->sf_tx_ctrl = htole32(SF_TX_DESC_ID |
(txsegs[i].ds_len & SF_TX_DESC_FRAGLEN));
desc->sf_tx_reserved = 0;
desc->sf_addr = htole64(txsegs[i].ds_addr);
if (i == 0 && prod + nsegs >= SF_TX_DLIST_CNT) {
/* Queue wraps! */
desc->sf_tx_ctrl |= htole32(SF_TX_DESC_END);
prod = 0;
} else
SF_INC(prod, SF_TX_DLIST_CNT);
}
/* Update producer index. */
sc->sf_cdata.sf_tx_prod = prod;
sc->sf_cdata.sf_tx_cnt += nsegs + nskip;
desc = &sc->sf_rdata.sf_tx_ring[si];
/* Check TDP/UDP checksum offload request. */
if ((m->m_pkthdr.csum_flags & SF_CSUM_FEATURES) != 0)
desc->sf_tx_ctrl |= htole32(SF_TX_DESC_CALTCP);
desc->sf_tx_ctrl |=
htole32(SF_TX_DESC_CRCEN | SF_TX_DESC_INTR | (nsegs << 16));
txd->tx_dmamap = map;
txd->tx_m = m;
txd->ndesc = nsegs + nskip;
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_start(struct ifnet *ifp)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct sf_softc *sc;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
sc = ifp->if_softc;
SF_LOCK(sc);
sf_start_locked(ifp);
SF_UNLOCK(sc);
}
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_start_locked(struct ifnet *ifp)
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
{
struct sf_softc *sc;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct mbuf *m_head;
int enq;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sc = ifp->if_softc;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK_ASSERT(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
IFF_DRV_RUNNING || sc->sf_link == 0)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
return;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* Since we don't know when descriptor wrap occurrs in advance
* limit available number of active Tx descriptor counter to be
* higher than maximum number of DMA segments allowed in driver.
*/
for (enq = 0; !IFQ_DRV_IS_EMPTY(&ifp->if_snd) &&
sc->sf_cdata.sf_tx_cnt < SF_TX_DLIST_CNT - SF_MAXTXSEGS; ) {
Add altq support. Patch by mlaier. Approved by: mlaier MFC after: 2 weeks
2004-11-12 18:12:04 +00:00
IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
if (m_head == NULL)
break;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* Pack the data into the transmit ring. If we
* don't have room, set the OACTIVE flag and wait
* for the NIC to drain the ring.
*/
if (sf_encap(sc, &m_head)) {
if (m_head == NULL)
break;
Add altq support. Patch by mlaier. Approved by: mlaier MFC after: 2 weeks
2004-11-12 18:12:04 +00:00
IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Check the return value of sf_encap() and handle errors accordingly.
2001-01-20 00:50:51 +00:00
break;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
enq++;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
/*
* If there's a BPF listener, bounce a copy of this frame
* to him.
*/
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ETHER_BPF_MTAP(ifp, m_head);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (enq > 0) {
bus_dmamap_sync(sc->sf_cdata.sf_tx_ring_tag,
sc->sf_cdata.sf_tx_ring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
/* Kick transmit. */
csr_write_4(sc, SF_TXDQ_PRODIDX,
sc->sf_cdata.sf_tx_prod * (sizeof(struct sf_tx_rdesc) / 8));
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Set a timeout in case the chip goes out to lunch. */
sc->sf_watchdog_timer = 5;
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_stop(struct sf_softc *sc)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_txdesc *txd;
struct sf_rxdesc *rxd;
Remember to clear the IFF_RUNNING and IFF_OACTIVE flags in sf_stop() and sk_stop().
1999-07-25 05:16:05 +00:00
struct ifnet *ifp;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
int i;
Remember to clear the IFF_RUNNING and IFF_OACTIVE flags in sf_stop() and sk_stop().
1999-07-25 05:16:05 +00:00
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK_ASSERT(sc);
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->sf_ifp;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
sc->sf_link = 0;
callout_stop(&sc->sf_co);
sc->sf_watchdog_timer = 0;
/* Reading the ISR register clears all interrrupts. */
csr_read_4(sc, SF_ISR);
/* Disable further interrupts. */
csr_write_4(sc, SF_IMR, 0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Disable Tx/Rx egine. */
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_GEN_ETH_CTL, 0);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
Give hardware chance to drain active DMA cycles.
2012-02-23 05:10:00 +00:00
/* Give hardware chance to drain active DMA cycles. */
DELAY(1000);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
csr_write_4(sc, SF_CQ_CONSIDX, 0);
csr_write_4(sc, SF_CQ_PRODIDX, 0);
csr_write_4(sc, SF_RXDQ_ADDR_Q1, 0);
csr_write_4(sc, SF_RXDQ_CTL_1, 0);
csr_write_4(sc, SF_RXDQ_PTR_Q1, 0);
csr_write_4(sc, SF_TXCQ_CTL, 0);
csr_write_4(sc, SF_TXDQ_ADDR_HIPRIO, 0);
csr_write_4(sc, SF_TXDQ_CTL, 0);
Modify the Adaptec "starfire" driver to reset the PHY on the MII bus before selecting a mode. The Seeq PHY chips on the Adaptec cards that use the AIC-6915 controller seem to need it in order to get them to change modes correctly. This corrects a problem that I noticed where my ANA-62022 board failed to correctly program the full duplex bit in the macconfig1 register when the interface was brought up. Running ifconfig sf0 would mask this problem in some cases because polling the PHY status would cause the miibus code to notice that full duplex was now needed and the statchg callback would be invoked to configure the duplex setting. However it would still get it wrong other times. Also changed sf_miibus_statchg() to program the IPG register to match the duplex setting in accordance with Adaptec manual's recommendations (0x15 for full duplex, 0x11 for half duplex).
1999-12-05 20:02:45 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/*
* Free RX and TX mbufs still in the queues.
*/
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
for (i = 0; i < SF_RX_DLIST_CNT; i++) {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
rxd = &sc->sf_cdata.sf_rxdesc[i];
if (rxd->rx_m != NULL) {
bus_dmamap_sync(sc->sf_cdata.sf_rx_tag,
rxd->rx_dmamap, BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->sf_cdata.sf_rx_tag,
rxd->rx_dmamap);
m_freem(rxd->rx_m);
rxd->rx_m = NULL;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
for (i = 0; i < SF_TX_DLIST_CNT; i++) {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
txd = &sc->sf_cdata.sf_txdesc[i];
if (txd->tx_m != NULL) {
bus_dmamap_sync(sc->sf_cdata.sf_tx_tag,
txd->tx_dmamap, BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(sc->sf_cdata.sf_tx_tag,
txd->tx_dmamap);
m_freem(txd->tx_m);
txd->tx_m = NULL;
txd->ndesc = 0;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
}
}
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static void
sf_tick(void *xsc)
{
struct sf_softc *sc;
struct mii_data *mii;
sc = xsc;
SF_LOCK_ASSERT(sc);
mii = device_get_softc(sc->sf_miibus);
mii_tick(mii);
sf_stats_update(sc);
sf_watchdog(sc);
callout_reset(&sc->sf_co, hz, sf_tick, sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
/*
* Note: it is important that this function not be interrupted. We
* use a two-stage register access scheme: if we are interrupted in
* between setting the indirect address register and reading from the
* indirect data register, the contents of the address register could
* be changed out from under us.
Whitespace nit
2004-07-05 22:52:01 +00:00
*/
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_stats_update(struct sf_softc *sc)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct ifnet *ifp;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct sf_stats now, *stats, *nstats;
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
int i;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK_ASSERT(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ifp = sc->sf_ifp;
stats = &now;
stats->sf_tx_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_FRAMES);
stats->sf_tx_single_colls =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_SINGLE_COL);
stats->sf_tx_multi_colls =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_MULTI_COL);
stats->sf_tx_crcerrs =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_CRC_ERRS);
stats->sf_tx_bytes =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_BYTES);
stats->sf_tx_deferred =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_DEFERRED);
stats->sf_tx_late_colls =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_LATE_COL);
stats->sf_tx_pause_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_PAUSE);
stats->sf_tx_control_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_CTL_FRAME);
stats->sf_tx_excess_colls =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_EXCESS_COL);
stats->sf_tx_excess_defer =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_EXCESS_DEF);
stats->sf_tx_mcast_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_MULTI);
stats->sf_tx_bcast_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_BCAST);
stats->sf_tx_frames_lost =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_FRAME_LOST);
stats->sf_rx_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_FRAMES);
stats->sf_rx_crcerrs =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_CRC_ERRS);
stats->sf_rx_alignerrs =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_ALIGN_ERRS);
stats->sf_rx_bytes =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_BYTES);
stats->sf_rx_pause_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_PAUSE);
stats->sf_rx_control_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_CTL_FRAME);
stats->sf_rx_unsup_control_frames =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_UNSUP_FRAME);
stats->sf_rx_giants =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_GIANTS);
stats->sf_rx_runts =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_RUNTS);
stats->sf_rx_jabbererrs =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_JABBER);
stats->sf_rx_fragments =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_FRAGMENTS);
stats->sf_rx_pkts_64 =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_64);
stats->sf_rx_pkts_65_127 =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_65_127);
stats->sf_rx_pkts_128_255 =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_128_255);
stats->sf_rx_pkts_256_511 =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_256_511);
stats->sf_rx_pkts_512_1023 =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_512_1023);
stats->sf_rx_pkts_1024_1518 =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_1024_1518);
stats->sf_rx_frames_lost =
csr_read_4(sc, SF_STATS_BASE + SF_STATS_RX_FRAME_LOST);
/* Lower 16bits are valid. */
stats->sf_tx_underruns =
(csr_read_4(sc, SF_STATS_BASE + SF_STATS_TX_UNDERRUN) & 0xffff);
Do not call mii_pollstat() from within device tick routines; the status information is updated by mii_tick(). Pointed out by: wpaul (a while back)
2001-09-29 19:28:31 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Empty stats counter registers. */
for (i = SF_STATS_BASE; i < (SF_STATS_END + 1); i += sizeof(uint32_t))
csr_write_4(sc, i, 0);
Convert the Adaptec and Winbond drivers to miibus.
1999-08-30 23:08:32 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
ifp->if_opackets += (u_long)stats->sf_tx_frames;
ifp->if_collisions += (u_long)stats->sf_tx_single_colls +
(u_long)stats->sf_tx_multi_colls;
ifp->if_oerrors += (u_long)stats->sf_tx_excess_colls +
(u_long)stats->sf_tx_excess_defer +
(u_long)stats->sf_tx_frames_lost;
ifp->if_ipackets += (u_long)stats->sf_rx_frames;
ifp->if_ierrors += (u_long)stats->sf_rx_crcerrs +
(u_long)stats->sf_rx_alignerrs +
(u_long)stats->sf_rx_giants +
(u_long)stats->sf_rx_runts +
(u_long)stats->sf_rx_jabbererrs +
(u_long)stats->sf_rx_frames_lost;
nstats = &sc->sf_statistics;
nstats->sf_tx_frames += stats->sf_tx_frames;
nstats->sf_tx_single_colls += stats->sf_tx_single_colls;
nstats->sf_tx_multi_colls += stats->sf_tx_multi_colls;
nstats->sf_tx_crcerrs += stats->sf_tx_crcerrs;
nstats->sf_tx_bytes += stats->sf_tx_bytes;
nstats->sf_tx_deferred += stats->sf_tx_deferred;
nstats->sf_tx_late_colls += stats->sf_tx_late_colls;
nstats->sf_tx_pause_frames += stats->sf_tx_pause_frames;
nstats->sf_tx_control_frames += stats->sf_tx_control_frames;
nstats->sf_tx_excess_colls += stats->sf_tx_excess_colls;
nstats->sf_tx_excess_defer += stats->sf_tx_excess_defer;
nstats->sf_tx_mcast_frames += stats->sf_tx_mcast_frames;
nstats->sf_tx_bcast_frames += stats->sf_tx_bcast_frames;
nstats->sf_tx_frames_lost += stats->sf_tx_frames_lost;
nstats->sf_rx_frames += stats->sf_rx_frames;
nstats->sf_rx_crcerrs += stats->sf_rx_crcerrs;
nstats->sf_rx_alignerrs += stats->sf_rx_alignerrs;
nstats->sf_rx_bytes += stats->sf_rx_bytes;
nstats->sf_rx_pause_frames += stats->sf_rx_pause_frames;
nstats->sf_rx_control_frames += stats->sf_rx_control_frames;
nstats->sf_rx_unsup_control_frames += stats->sf_rx_unsup_control_frames;
nstats->sf_rx_giants += stats->sf_rx_giants;
nstats->sf_rx_runts += stats->sf_rx_runts;
nstats->sf_rx_jabbererrs += stats->sf_rx_jabbererrs;
nstats->sf_rx_fragments += stats->sf_rx_fragments;
nstats->sf_rx_pkts_64 += stats->sf_rx_pkts_64;
nstats->sf_rx_pkts_65_127 += stats->sf_rx_pkts_65_127;
nstats->sf_rx_pkts_128_255 += stats->sf_rx_pkts_128_255;
nstats->sf_rx_pkts_256_511 += stats->sf_rx_pkts_256_511;
nstats->sf_rx_pkts_512_1023 += stats->sf_rx_pkts_512_1023;
nstats->sf_rx_pkts_1024_1518 += stats->sf_rx_pkts_1024_1518;
nstats->sf_rx_frames_lost += stats->sf_rx_frames_lost;
nstats->sf_tx_underruns += stats->sf_tx_underruns;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
style: put return values on a line by themselves. fix some paste issues where whitespace was used instead of tabs.
2002-08-23 23:49:02 +00:00
static void
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_watchdog(struct sf_softc *sc)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
struct ifnet *ifp;
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
SF_LOCK_ASSERT(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_watchdog_timer == 0 || --sc->sf_watchdog_timer)
return;
ifp = sc->sf_ifp;
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
ifp->if_oerrors++;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
if (sc->sf_link == 0) {
if (bootverbose)
if_printf(sc->sf_ifp, "watchdog timeout "
"(missed link)\n");
} else
if_printf(ifp, "watchdog timeout, %d Tx descs are active\n",
sc->sf_cdata.sf_tx_cnt);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Remove unnecessary controller reinitialization. StarFire controller does not require controller reinitialization to program perfect filters. While here, make driver immediately exit from interrupt/polling handler if driver reinitialized controller. PR: kern/87506
2010-09-21 17:42:19 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
sf_init_locked(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Add altq support. Patch by mlaier. Approved by: mlaier MFC after: 2 weeks
2004-11-12 18:12:04 +00:00
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
sf_start_locked(ifp);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
}
static int
sf_shutdown(device_t dev)
{
struct sf_softc *sc;
sc = device_get_softc(dev);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
SF_LOCK(sc);
sf_stop(sc);
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
SF_UNLOCK(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Fix function prototype for device_shutdown method.
2007-11-22 02:45:00 +00:00
static int
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sf_suspend(device_t dev)
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
{
struct sf_softc *sc;
sc = device_get_softc(dev);
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_LOCK(sc);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
sf_stop(sc);
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
sc->sf_suspended = 1;
bus_generic_suspend(dev);
SF_UNLOCK(sc);
return (0);
}
static int
sf_resume(device_t dev)
{
struct sf_softc *sc;
struct ifnet *ifp;
sc = device_get_softc(dev);
SF_LOCK(sc);
bus_generic_resume(dev);
ifp = sc->sf_ifp;
if ((ifp->if_flags & IFF_UP) != 0)
sf_init_locked(sc);
sc->sf_suspended = 0;
Fixup locking for sf(4) and mark MPSAFE: - Add locked variants of start, init, and ifmedia_upd. - Use callout_* instead of timeout/untimeout. - Don't recurse on the driver lock. - Fixup locking in ioctl. - Lock the driver lock in the ifmedia handlers rather than across ifmedia_ioctl(). Tested by: brueffer MFC after: 3 days
2005-08-18 17:09:16 +00:00
SF_UNLOCK(sc);
Fix function prototype for device_shutdown method.
2007-11-22 02:45:00 +00:00
return (0);
This commit adds device driver support for Adaptec Duralink PCI fast ethernet controllers based on the AIC-6915 "Starfire" controller chip. There are single port, dual port and quad port cards, plus one 100baseFX card. All are 64-bit PCI devices, except one single port model. The Starfire would be a very nice chip were it not for the fact that receive buffers have to be longword aligned. This requires buffer copying in order to achieve proper payload alignment on the alpha. Payload alignment is enforced on both the alpha and x86 platforms. The Starfire has several different DMA descriptor formats and transfer mechanisms. This driver uses frame descriptors for transmission which can address up to 14 packet fragments, and a single fragment descriptor for receive. It also uses the producer/consumer model and completion queues for both transmit and receive. The transmit ring has 128 descriptors and the receive ring has 256. This driver supports both FreeBSD/i386 and FreeBSD/alpha, and uses newbus so that it can be compiled as a loadable kernel module. Support for BPF and hardware multicast filtering is included.
1999-07-25 04:32:50 +00:00
}
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
static int
sf_sysctl_stats(SYSCTL_HANDLER_ARGS)
{
struct sf_softc *sc;
struct sf_stats *stats;
int error;
int result;
result = -1;
error = sysctl_handle_int(oidp, &result, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if (result != 1)
return (error);
sc = (struct sf_softc *)arg1;
stats = &sc->sf_statistics;
printf("%s statistics:\n", device_get_nameunit(sc->sf_dev));
printf("Transmit good frames : %ju\n",
(uintmax_t)stats->sf_tx_frames);
printf("Transmit good octets : %ju\n",
(uintmax_t)stats->sf_tx_bytes);
printf("Transmit single collisions : %u\n",
stats->sf_tx_single_colls);
printf("Transmit multiple collisions : %u\n",
stats->sf_tx_multi_colls);
printf("Transmit late collisions : %u\n",
stats->sf_tx_late_colls);
printf("Transmit abort due to excessive collisions : %u\n",
stats->sf_tx_excess_colls);
printf("Transmit CRC errors : %u\n",
stats->sf_tx_crcerrs);
printf("Transmit deferrals : %u\n",
stats->sf_tx_deferred);
printf("Transmit abort due to excessive deferrals : %u\n",
stats->sf_tx_excess_defer);
printf("Transmit pause control frames : %u\n",
stats->sf_tx_pause_frames);
printf("Transmit control frames : %u\n",
stats->sf_tx_control_frames);
printf("Transmit good multicast frames : %u\n",
stats->sf_tx_mcast_frames);
printf("Transmit good broadcast frames : %u\n",
stats->sf_tx_bcast_frames);
printf("Transmit frames lost due to internal transmit errors : %u\n",
stats->sf_tx_frames_lost);
printf("Transmit FIFO underflows : %u\n",
stats->sf_tx_underruns);
printf("Transmit GFP stalls : %u\n", stats->sf_tx_gfp_stall);
printf("Receive good frames : %ju\n",
(uint64_t)stats->sf_rx_frames);
printf("Receive good octets : %ju\n",
(uint64_t)stats->sf_rx_bytes);
printf("Receive CRC errors : %u\n",
stats->sf_rx_crcerrs);
printf("Receive alignment errors : %u\n",
stats->sf_rx_alignerrs);
printf("Receive pause frames : %u\n",
stats->sf_rx_pause_frames);
printf("Receive control frames : %u\n",
stats->sf_rx_control_frames);
printf("Receive control frames with unsupported opcode : %u\n",
stats->sf_rx_unsup_control_frames);
printf("Receive frames too long : %u\n",
stats->sf_rx_giants);
printf("Receive frames too short : %u\n",
stats->sf_rx_runts);
printf("Receive frames jabber errors : %u\n",
stats->sf_rx_jabbererrs);
printf("Receive frames fragments : %u\n",
stats->sf_rx_fragments);
printf("Receive packets 64 bytes : %ju\n",
(uint64_t)stats->sf_rx_pkts_64);
printf("Receive packets 65 to 127 bytes : %ju\n",
(uint64_t)stats->sf_rx_pkts_65_127);
printf("Receive packets 128 to 255 bytes : %ju\n",
(uint64_t)stats->sf_rx_pkts_128_255);
printf("Receive packets 256 to 511 bytes : %ju\n",
(uint64_t)stats->sf_rx_pkts_256_511);
printf("Receive packets 512 to 1023 bytes : %ju\n",
(uint64_t)stats->sf_rx_pkts_512_1023);
printf("Receive packets 1024 to 1518 bytes : %ju\n",
(uint64_t)stats->sf_rx_pkts_1024_1518);
printf("Receive frames lost due to internal receive errors : %u\n",
stats->sf_rx_frames_lost);
printf("Receive GFP stalls : %u\n", stats->sf_rx_gfp_stall);
return (error);
}
static int
sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high)
{
int error, value;
if (!arg1)
return (EINVAL);
value = *(int *)arg1;
error = sysctl_handle_int(oidp, &value, 0, req);
if (error || !req->newptr)
return (error);
if (value < low || value > high)
return (EINVAL);
*(int *)arg1 = value;
return (0);
}
static int
sysctl_hw_sf_int_mod(SYSCTL_HANDLER_ARGS)
{
return (sysctl_int_range(oidp, arg1, arg2, req, SF_IM_MIN, SF_IM_MAX));
}
Reference in New Issue Copy Permalink