freebsd-skq/sys/dev/sf/if_sfreg.h

1105 lines
34 KiB
C
Raw Normal View History

/*-
* 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.
*
1999-08-28 01:08:13 +00:00
* $FreeBSD$
*/
/*
* Registers for the Adaptec AIC-6915 Starfire. The Starfire has a 512K
* register space. These registers can be accessed in the following way:
* - PCI config registers are always accessible through PCI config space
* - Full 512K space mapped into memory using PCI memory mapped access
* - 256-byte I/O space mapped through PCI I/O access
* - Full 512K space mapped through indirect I/O using PCI I/O access
* It's possible to use either memory mapped mode or I/O mode to access
* the registers, but memory mapped is usually the easiest. All registers
* are 32 bits wide and must be accessed using 32-bit operations.
*/
/*
* Adaptec PCI vendor ID.
*/
#define AD_VENDORID 0x9004
/*
* AIC-6915 PCI device ID.
*/
#define AD_DEVICEID_STARFIRE 0x6915
/*
* AIC-6915 subsystem IDs. Adaptec uses the subsystem ID to identify
* the exact kind of NIC on which the ASIC is mounted. Currently there
* are six different variations. Note: the Adaptec manual lists code 0x28
* for two different NICs: the 62044 and the 69011/TX. This is a typo:
* the code for the 62044 is really 0x18.
*
* Note that there also appears to be an 0x19 code for a newer rev
* 62044 card.
*/
#define AD_SUBSYSID_62011_REV0 0x0008 /* single port 10/100baseTX 64-bit */
#define AD_SUBSYSID_62011_REV1 0x0009 /* single port 10/100baseTX 64-bit */
#define AD_SUBSYSID_62022 0x0010 /* dual port 10/100baseTX 64-bit */
#define AD_SUBSYSID_62044_REV0 0x0018 /* quad port 10/100baseTX 64-bit */
#define AD_SUBSYSID_62044_REV1 0x0019 /* quad port 10/100baseTX 64-bit */
#define AD_SUBSYSID_62020 0x0020 /* single port 10/100baseFX 64-bit */
#define AD_SUBSYSID_69011 0x0028 /* single port 10/100baseTX 32-bit */
/*
* Starfire internal register space map. The entire register space
* is available using PCI memory mapped mode. The SF_RMAP_INTREG
* space is available using PCI I/O mode. The entire space can be
* accessed using indirect I/O using the indirect I/O addr and
* indirect I/O data registers located within the SF_RMAP_INTREG space.
*/
#define SF_RMAP_ROMADDR_BASE 0x00000 /* Expansion ROM space */
#define SF_RMAP_ROMADDR_MAX 0x3FFFF
#define SF_RMAP_EXGPIO_BASE 0x40000 /* External general purpose regs */
#define SF_RMAP_EXGPIO_MAX 0x3FFFF
#define SF_RMAP_INTREG_BASE 0x50000 /* Internal functional registers */
#define SF_RMAP_INTREG_MAX 0x500FF
#define SF_RMAP_GENREG_BASE 0x50100 /* General purpose registers */
#define SF_RMAP_GENREG_MAX 0x5FFFF
#define SF_RMAP_FIFO_BASE 0x60000
#define SF_RMAP_FIFO_MAX 0x6FFFF
#define SF_RMAP_STS_BASE 0x70000
#define SF_RMAP_STS_MAX 0x70083
#define SF_RMAP_RSVD_BASE 0x70084
#define SF_RMAP_RSVD_MAX 0x7FFFF
/*
* PCI config header registers, 0x0000 to 0x003F
*/
#define SF_PCI_VENDOR_ID 0x0000
#define SF_PCI_DEVICE_ID 0x0002
#define SF_PCI_COMMAND 0x0004
#define SF_PCI_STATUS 0x0006
#define SF_PCI_REVID 0x0008
#define SF_PCI_CLASSCODE 0x0009
#define SF_PCI_CACHELEN 0x000C
#define SF_PCI_LATENCY_TIMER 0x000D
#define SF_PCI_HEADER_TYPE 0x000E
#define SF_PCI_LOMEM 0x0010
#define SF_PCI_LOIO 0x0014
#define SF_PCI_SUBVEN_ID 0x002C
#define SF_PCI_SYBSYS_ID 0x002E
#define SF_PCI_BIOSROM 0x0030
#define SF_PCI_INTLINE 0x003C
#define SF_PCI_INTPIN 0x003D
#define SF_PCI_MINGNT 0x003E
#define SF_PCI_MINLAT 0x003F
/*
* PCI registers, 0x0040 to 0x006F
*/
#define SF_PCI_DEVCFG 0x0040
#define SF_BACCTL 0x0044
#define SF_PCI_MON1 0x0048
#define SF_PCI_MON2 0x004C
#define SF_PCI_CAPID 0x0050 /* 8 bits */
#define SF_PCI_NEXTPTR 0x0051 /* 8 bits */
#define SF_PCI_PWRMGMTCAP 0x0052 /* 16 bits */
#define SF_PCI_PWRMGMTCTRL 0x0054 /* 16 bits */
#define SF_PCI_PME_EVENT 0x0058
#define SF_PCI_EECTL 0x0060
#define SF_PCI_COMPLIANCE 0x0064
#define SF_INDIRECTIO_ADDR 0x0068
#define SF_INDIRECTIO_DATA 0x006C
#define SF_PCIDEVCFG_RESET 0x00000001
#define SF_PCIDEVCFG_FORCE64 0x00000002
#define SF_PCIDEVCFG_SYSTEM64 0x00000004
#define SF_PCIDEVCFG_RSVD0 0x00000008
#define SF_PCIDEVCFG_INCR_INB 0x00000010
#define SF_PCIDEVCFG_ABTONPERR 0x00000020
#define SF_PCIDEVCFG_STPONPERR 0x00000040
#define SF_PCIDEVCFG_MR_ENB 0x00000080
#define SF_PCIDEVCFG_FIFOTHR 0x00000F00
#define SF_PCIDEVCFG_STPONCA 0x00001000
#define SF_PCIDEVCFG_PCIMEN 0x00002000 /* enable PCI bus master */
#define SF_PCIDEVCFG_LATSTP 0x00004000
#define SF_PCIDEVCFG_BYTE_ENB 0x00008000
#define SF_PCIDEVCFG_EECSWIDTH 0x00070000
#define SF_PCIDEVCFG_STPMWCA 0x00080000
#define SF_PCIDEVCFG_REGCSWIDTH 0x00700000
#define SF_PCIDEVCFG_INTR_ENB 0x00800000
#define SF_PCIDEVCFG_DPR_ENB 0x01000000
#define SF_PCIDEVCFG_RSVD1 0x02000000
#define SF_PCIDEVCFG_RSVD2 0x04000000
#define SF_PCIDEVCFG_STA_ENB 0x08000000
#define SF_PCIDEVCFG_RTA_ENB 0x10000000
#define SF_PCIDEVCFG_RMA_ENB 0x20000000
#define SF_PCIDEVCFG_SSE_ENB 0x40000000
#define SF_PCIDEVCFG_DPE_ENB 0x80000000
#define SF_BACCTL_BACDMA_ENB 0x00000001
#define SF_BACCTL_PREFER_RXDMA 0x00000002
#define SF_BACCTL_PREFER_TXDMA 0x00000004
#define SF_BACCTL_SINGLE_DMA 0x00000008
#define SF_BACCTL_SWAPMODE_DATA 0x00000030
#define SF_BACCTL_SWAPMODE_DESC 0x000000C0
#define SF_SWAPMODE_LE 0x00000000
#define SF_SWAPMODE_BE 0x00000010
#define SF_PSTATE_MASK 0x0003
#define SF_PSTATE_D0 0x0000
#define SF_PSTATE_D1 0x0001
#define SF_PSTATE_D2 0x0002
#define SF_PSTATE_D3 0x0003
#define SF_PME_EN 0x0010
#define SF_PME_STATUS 0x8000
/*
* Ethernet registers 0x0070 to 0x00FF
*/
#define SF_GEN_ETH_CTL 0x0070
#define SF_TIMER_CTL 0x0074
#define SF_CURTIME 0x0078
#define SF_ISR 0x0080
#define SF_ISR_SHADOW 0x0084
#define SF_IMR 0x0088
#define SF_GPIO 0x008C
#define SF_TXDQ_CTL 0x0090
#define SF_TXDQ_ADDR_HIPRIO 0x0094
#define SF_TXDQ_ADDR_LOPRIO 0x0098
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_TXDQ_ADDR_HI 0x009C
#define SF_TXDQ_PRODIDX 0x00A0
#define SF_TXDQ_CONSIDX 0x00A4
#define SF_TXDMA_STS1 0x00A8
#define SF_TXDMA_STS2 0x00AC
#define SF_TX_FRAMCTL 0x00B0
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_CQ_ADDR_HI 0x00B4
#define SF_TXCQ_CTL 0x00B8
#define SF_RXCQ_CTL_1 0x00BC
#define SF_RXCQ_CTL_2 0x00C0
#define SF_CQ_CONSIDX 0x00C4
#define SF_CQ_PRODIDX 0x00C8
#define SF_CQ_RXQ2 0x00CC
#define SF_RXDMA_CTL 0x00D0
#define SF_RXDQ_CTL_1 0x00D4
#define SF_RXDQ_CTL_2 0x00D8
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_RXDQ_ADDR_HI 0x00DC
#define SF_RXDQ_ADDR_Q1 0x00E0
#define SF_RXDQ_ADDR_Q2 0x00E4
#define SF_RXDQ_PTR_Q1 0x00E8
#define SF_RXDQ_PTR_Q2 0x00EC
#define SF_RXDMA_STS 0x00F0
#define SF_RXFILT 0x00F4
#define SF_RX_FRAMETEST_OUT 0x00F8
/* Ethernet control register */
#define SF_ETHCTL_RX_ENB 0x00000001
#define SF_ETHCTL_TX_ENB 0x00000002
#define SF_ETHCTL_RXDMA_ENB 0x00000004
#define SF_ETHCTL_TXDMA_ENB 0x00000008
#define SF_ETHCTL_RXGFP_ENB 0x00000010
#define SF_ETHCTL_TXGFP_ENB 0x00000020
#define SF_ETHCTL_SOFTINTR 0x00000800
/* Timer control register */
#define SF_TIMER_IMASK_INTERVAL 0x0000001F
#define SF_TIMER_IMASK_MODE 0x00000060
#define SF_TIMER_SMALLFRAME_BYP 0x00000100
#define SF_TIMER_SMALLRX_FRAME 0x00000600
#define SF_TIMER_TIMES_TEN 0x00000800
#define SF_TIMER_RXHIPRIO_BYP 0x00001000
#define SF_TIMER_TX_DMADONE_DLY 0x00002000
#define SF_TIMER_TX_QDONE_DLY 0x00004000
#define SF_TIMER_TX_FRDONE_DLY 0x00008000
#define SF_TIMER_GENTIMER 0x00FF0000
#define SF_TIMER_ONESHOT 0x01000000
#define SF_TIMER_GENTIMER_RES 0x02000000
#define SF_TIMER_TIMEST_RES 0x04000000
#define SF_TIMER_RXQ2DONE_DLY 0x10000000
#define SF_TIMER_EARLYRX2_DLY 0x20000000
#define SF_TIMER_RXQ1DONE_DLY 0x40000000
#define SF_TIMER_EARLYRX1_DLY 0x80000000
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Timer resolution is 0.8us * 128. */
#define SF_IM_MIN 0
#define SF_IM_MAX 0x1F /* 3.276ms */
#define SF_IM_DEFAULT 1 /* 102.4us */
/* Interrupt status register */
#define SF_ISR_PCIINT_ASSERTED 0x00000001
#define SF_ISR_GFP_TX 0x00000002
#define SF_ISR_GFP_RX 0x00000004
#define SF_ISR_TX_BADID_HIPRIO 0x00000008
#define SF_ISR_TX_BADID_LOPRIO 0x00000010
#define SF_ISR_NO_TX_CSUM 0x00000020
#define SF_ISR_RXDQ2_NOBUFS 0x00000040
#define SF_ISR_RXGFP_NORESP 0x00000080
#define SF_ISR_RXDQ1_DMADONE 0x00000100
#define SF_ISR_RXDQ2_DMADONE 0x00000200
#define SF_ISR_RXDQ1_EARLY 0x00000400
#define SF_ISR_RXDQ2_EARLY 0x00000800
#define SF_ISR_TX_QUEUEDONE 0x00001000
#define SF_ISR_TX_DMADONE 0x00002000
#define SF_ISR_TX_TXDONE 0x00004000
#define SF_ISR_NORMALINTR 0x00008000
#define SF_ISR_RXDQ1_NOBUFS 0x00010000
#define SF_ISR_RXCQ2_NOBUFS 0x00020000
#define SF_ISR_TX_LOFIFO 0x00040000
#define SF_ISR_DMAERR 0x00080000
#define SF_ISR_PCIINT 0x00100000
#define SF_ISR_TXCQ_NOBUFS 0x00200000
#define SF_ISR_RXCQ1_NOBUFS 0x00400000
#define SF_ISR_SOFTINTR 0x00800000
#define SF_ISR_GENTIMER 0x01000000
#define SF_ISR_ABNORMALINTR 0x02000000
#define SF_ISR_RSVD0 0x04000000
#define SF_ISR_STATSOFLOW 0x08000000
#define SF_ISR_GPIO 0xF0000000
/*
* Shadow interrupt status register. Unlike the normal IRQ register,
* reading bits here does not automatically cause them to reset.
*/
#define SF_SISR_PCIINT_ASSERTED 0x00000001
#define SF_SISR_GFP_TX 0x00000002
#define SF_SISR_GFP_RX 0x00000004
#define SF_SISR_TX_BADID_HIPRIO 0x00000008
#define SF_SISR_TX_BADID_LOPRIO 0x00000010
#define SF_SISR_NO_TX_CSUM 0x00000020
#define SF_SISR_RXDQ2_NOBUFS 0x00000040
#define SF_SISR_RXGFP_NORESP 0x00000080
#define SF_SISR_RXDQ1_DMADONE 0x00000100
#define SF_SISR_RXDQ2_DMADONE 0x00000200
#define SF_SISR_RXDQ1_EARLY 0x00000400
#define SF_SISR_RXDQ2_EARLY 0x00000800
#define SF_SISR_TX_QUEUEDONE 0x00001000
#define SF_SISR_TX_DMADONE 0x00002000
#define SF_SISR_TX_TXDONE 0x00004000
#define SF_SISR_NORMALINTR 0x00008000
#define SF_SISR_RXDQ1_NOBUFS 0x00010000
#define SF_SISR_RXCQ2_NOBUFS 0x00020000
#define SF_SISR_TX_LOFIFO 0x00040000
#define SF_SISR_DMAERR 0x00080000
#define SF_SISR_PCIINT 0x00100000
#define SF_SISR_TXCQ_NOBUFS 0x00200000
#define SF_SISR_RXCQ1_NOBUFS 0x00400000
#define SF_SISR_SOFTINTR 0x00800000
#define SF_SISR_GENTIMER 0x01000000
#define SF_SISR_ABNORMALINTR 0x02000000
#define SF_SISR_RSVD0 0x04000000
#define SF_SISR_STATSOFLOW 0x08000000
#define SF_SISR_GPIO 0xF0000000
/* Interrupt mask register */
#define SF_IMR_PCIINT_ASSERTED 0x00000001
#define SF_IMR_GFP_TX 0x00000002
#define SF_IMR_GFP_RX 0x00000004
#define SF_IMR_TX_BADID_HIPRIO 0x00000008
#define SF_IMR_TX_BADID_LOPRIO 0x00000010
#define SF_IMR_NO_TX_CSUM 0x00000020
#define SF_IMR_RXDQ2_NOBUFS 0x00000040
#define SF_IMR_RXGFP_NORESP 0x00000080
#define SF_IMR_RXDQ1_DMADONE 0x00000100
#define SF_IMR_RXDQ2_DMADONE 0x00000200
#define SF_IMR_RXDQ1_EARLY 0x00000400
#define SF_IMR_RXDQ2_EARLY 0x00000800
#define SF_IMR_TX_QUEUEDONE 0x00001000
#define SF_IMR_TX_DMADONE 0x00002000
#define SF_IMR_TX_TXDONE 0x00004000
#define SF_IMR_NORMALINTR 0x00008000
#define SF_IMR_RXDQ1_NOBUFS 0x00010000
#define SF_IMR_RXCQ2_NOBUFS 0x00020000
#define SF_IMR_TX_LOFIFO 0x00040000
#define SF_IMR_DMAERR 0x00080000
#define SF_IMR_PCIINT 0x00100000
#define SF_IMR_TXCQ_NOBUFS 0x00200000
#define SF_IMR_RXCQ1_NOBUFS 0x00400000
#define SF_IMR_SOFTINTR 0x00800000
#define SF_IMR_GENTIMER 0x01000000
#define SF_IMR_ABNORMALINTR 0x02000000
#define SF_IMR_RSVD0 0x04000000
#define SF_IMR_STATSOFLOW 0x08000000
#define SF_IMR_GPIO 0xF0000000
#define SF_INTRS \
(SF_IMR_RXDQ2_NOBUFS|SF_IMR_RXDQ1_DMADONE|SF_IMR_RXDQ2_DMADONE| \
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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_IMR_TX_DMADONE|SF_IMR_RXDQ1_NOBUFS|SF_IMR_RXDQ2_DMADONE| \
SF_IMR_NORMALINTR|SF_IMR_ABNORMALINTR|SF_IMR_TXCQ_NOBUFS| \
SF_IMR_RXCQ1_NOBUFS|SF_IMR_RXCQ2_NOBUFS|SF_IMR_STATSOFLOW| \
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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_IMR_TX_LOFIFO|SF_IMR_DMAERR|SF_IMR_RXGFP_NORESP| \
SF_IMR_NO_TX_CSUM)
/* TX descriptor queue control registers */
#define SF_TXDQCTL_DESCTYPE 0x00000007
#define SF_TXDQCTL_NODMACMP 0x00000008
#define SF_TXDQCTL_MINSPACE 0x00000070
#define SF_TXDQCTL_64BITADDR 0x00000080
#define SF_TXDQCTL_BURSTLEN 0x00003F00
#define SF_TXDQCTL_SKIPLEN 0x001F0000
#define SF_TXDQCTL_HIPRIOTHRESH 0xFF000000
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_TXDMA_HIPRIO_THRESH 2
#define SF_TXDDMA_BURST (128 / 32)
#define SF_TXBUFDESC_TYPE0 0x00000000
#define SF_TXBUFDESC_TYPE1 0x00000001
#define SF_TXBUFDESC_TYPE2 0x00000002
#define SF_TXBUFDESC_TYPE3 0x00000003
#define SF_TXBUFDESC_TYPE4 0x00000004
#define SF_TXMINSPACE_UNLIMIT 0x00000000
#define SF_TXMINSPACE_32BYTES 0x00000010
#define SF_TXMINSPACE_64BYTES 0x00000020
#define SF_TXMINSPACE_128BYTES 0x00000030
#define SF_TXMINSPACE_256BYTES 0x00000040
#define SF_TXSKIPLEN_0BYTES 0x00000000
#define SF_TXSKIPLEN_8BYTES 0x00010000
#define SF_TXSKIPLEN_16BYTES 0x00020000
#define SF_TXSKIPLEN_24BYTES 0x00030000
#define SF_TXSKIPLEN_32BYTES 0x00040000
/* TX frame control register */
#define SF_TXFRMCTL_TXTHRESH 0x000000FF
#define SF_TXFRMCTL_CPLAFTERTX 0x00000100
#define SF_TXFRMCRL_DEBUG 0x0000FE00
#define SF_TXFRMCTL_STATUS 0x01FF0000
#define SF_TXFRMCTL_MAC_TXIF 0xFE000000
/* TX completion queue control register */
#define SF_TXCQ_THRESH 0x0000000F
#define SF_TXCQ_COMMON 0x00000010
#define SF_TXCQ_SIZE 0x00000020
#define SF_TXCQ_WRITEENB 0x00000040
#define SF_TXCQ_USE_64BIT 0x00000080
#define SF_TXCQ_ADDR 0xFFFFFF00
/* RX completion queue control register */
#define SF_RXCQ_THRESH 0x0000000F
#define SF_RXCQ_TYPE 0x00000030
#define SF_RXCQ_WRITEENB 0x00000040
#define SF_RXCQ_USE_64BIT 0x00000080
#define SF_RXCQ_ADDR 0xFFFFFF00
#define SF_RXCQTYPE_0 0x00000000
#define SF_RXCQTYPE_1 0x00000010
#define SF_RXCQTYPE_2 0x00000020
#define SF_RXCQTYPE_3 0x00000030
/* TX descriptor queue producer index register */
#define SF_TXDQ_PRODIDX_LOPRIO 0x000007FF
#define SF_TXDQ_PRODIDX_HIPRIO 0x07FF0000
/* TX descriptor queue consumer index register */
#define SF_TXDQ_CONSIDX_LOPRIO 0x000007FF
#define SF_TXDQ_CONSIDX_HIPRIO 0x07FF0000
/* Completion queue consumer index register */
#define SF_CQ_CONSIDX_RXQ1 0x000003FF
#define SF_CQ_CONSIDX_RXTHRMODE 0x00008000
#define SF_CQ_CONSIDX_TXQ 0x03FF0000
#define SF_CQ_CONSIDX_TXTHRMODE 0x80000000
/* Completion queue producer index register */
#define SF_CQ_PRODIDX_RXQ1 0x000003FF
#define SF_CQ_PRODIDX_TXQ 0x03FF0000
/* RX completion queue 2 consumer/producer index register */
#define SF_CQ_RXQ2_CONSIDX 0x000003FF
#define SF_CQ_RXQ2_RXTHRMODE 0x00008000
#define SF_CQ_RXQ2_PRODIDX 0x03FF0000
#define SF_CQ_RXTHRMODE_INT_ON 0x00008000
#define SF_CQ_RXTHRMODE_INT_OFF 0x00000000
#define SF_CQ_TXTHRMODE_INT_ON 0x80000000
#define SF_CQ_TXTHRMODE_INT_OFF 0x00000000
/* RX DMA control register */
#define SF_RXDMA_BURSTSIZE 0x0000007F
#define SF_RXDMA_FPTESTMODE 0x00000080
#define SF_RXDMA_HIPRIOTHRESH 0x00000F00
#define SF_RXDMA_RXEARLYTHRESH 0x0001F000
#define SF_RXDMA_DMACRC 0x00040000
#define SF_RXDMA_USEBKUPQUEUE 0x00080000
#define SF_RXDMA_QUEUEMODE 0x00700000
#define SF_RXDMA_RXCQ2_ON 0x00800000
#define SF_RXDMA_CSUMMODE 0x03000000
#define SF_RXDMA_DMAPAUSEPKTS 0x04000000
#define SF_RXDMA_DMACTLPKTS 0x08000000
#define SF_RXDMA_DMACRXERRPKTS 0x10000000
#define SF_RXDMA_DMABADPKTS 0x20000000
#define SF_RXDMA_DMARUNTS 0x40000000
#define SF_RXDMA_REPORTBADPKTS 0x80000000
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_RXDMA_HIGHPRIO_THRESH 6
#define SF_RXDMA_BURST (64 / 32)
#define SF_RXDQMODE_Q1ONLY 0x00100000
#define SF_RXDQMODE_Q2_ON_FP 0x00200000
#define SF_RXDQMODE_Q2_ON_SHORT 0x00300000
#define SF_RXDQMODE_Q2_ON_PRIO 0x00400000
#define SF_RXDQMODE_SPLITHDR 0x00500000
#define SF_RXCSUMMODE_IGNORE 0x00000000
#define SF_RXCSUMMODE_REJECT_BAD_TCP 0x01000000
#define SF_RXCSUMMODE_REJECT_BAD_TCPUDP 0x02000000
#define SF_RXCSUMMODE_RSVD 0x03000000
/* RX descriptor queue control registers */
#define SF_RXDQCTL_MINDESCTHR 0x0000007F
#define SF_RXDQCTL_Q1_WE 0x00000080
#define SF_RXDQCTL_DESCSPACE 0x00000700
#define SF_RXDQCTL_64BITDADDR 0x00000800
#define SF_RXDQCTL_64BITBADDR 0x00001000
#define SF_RXDQCTL_VARIABLE 0x00002000
#define SF_RXDQCTL_ENTRIES 0x00004000
#define SF_RXDQCTL_PREFETCH 0x00008000
#define SF_RXDQCTL_BUFLEN 0xFFFF0000
#define SF_DESCSPACE_4BYTES 0x00000000
#define SF_DESCSPACE_8BYTES 0x00000100
#define SF_DESCSPACE_16BYTES 0x00000200
#define SF_DESCSPACE_32BYTES 0x00000300
#define SF_DESCSPACE_64BYTES 0x00000400
#define SF_DESCSPACE_128_BYTES 0x00000500
/* RX buffer consumer/producer index registers */
#define SF_RXDQ_PRODIDX 0x000007FF
#define SF_RXDQ_CONSIDX 0x07FF0000
/* RX filter control register */
#define SF_RXFILT_PROMISC 0x00000001
#define SF_RXFILT_ALLMULTI 0x00000002
#define SF_RXFILT_BROAD 0x00000004
#define SF_RXFILT_HASHPRIO 0x00000008
#define SF_RXFILT_HASHMODE 0x00000030
#define SF_RXFILT_PERFMODE 0x000000C0
#define SF_RXFILT_VLANMODE 0x00000300
#define SF_RXFILT_WAKEMODE 0x00000C00
#define SF_RXFILT_MULTI_NOBROAD 0x00001000
#define SF_RXFILT_MIN_VLANPRIO 0x0000E000
#define SF_RXFILT_PEFECTPRIO 0xFFFF0000
/* Hash filtering mode */
#define SF_HASHMODE_OFF 0x00000000
#define SF_HASHMODE_WITHVLAN 0x00000010
#define SF_HASHMODE_ANYVLAN 0x00000020
#define SF_HASHMODE_ANY 0x00000030
/* Perfect filtering mode */
#define SF_PERFMODE_OFF 0x00000000
#define SF_PERFMODE_NORMAL 0x00000040
#define SF_PERFMODE_INVERSE 0x00000080
#define SF_PERFMODE_VLAN 0x000000C0
/* VLAN mode */
#define SF_VLANMODE_OFF 0x00000000
#define SF_VLANMODE_NOSTRIP 0x00000100
#define SF_VLANMODE_STRIP 0x00000200
#define SF_VLANMODE_RSVD 0x00000300
/* Wakeup mode */
#define SF_WAKEMODE_OFF 0x00000000
#define SF_WAKEMODE_FILTER 0x00000400
#define SF_WAKEMODE_FP 0x00000800
#define SF_WAKEMODE_HIPRIO 0x00000C00
/*
* Extra PCI registers 0x0100 to 0x0FFF
*/
#define SF_PCI_TARGSTAT 0x0100
#define SF_PCI_MASTSTAT1 0x0104
#define SF_PCI_MASTSTAT2 0x0108
#define SF_PCI_DMAHOSTADDR_LO 0x010C
#define SF_BAC_DMADIAG0 0x0110
#define SF_BAC_DMADIAG1 0x0114
#define SF_BAC_DMADIAG2 0x0118
#define SF_BAC_DMADIAG3 0x011C
#define SF_PAR0 0x0120
#define SF_PAR1 0x0124
#define SF_PCICB_FUNCEVENT 0x0130
#define SF_PCICB_FUNCEVENT_MASK 0x0134
#define SF_PCICB_FUNCSTATE 0x0138
#define SF_PCICB_FUNCFORCE 0x013C
/*
* Serial EEPROM registers 0x1000 to 0x1FFF
* Presumeably the EEPROM is mapped into this 8K window.
*/
#define SF_EEADDR_BASE 0x1000
#define SF_EEADDR_MAX 0x1FFF
#define SF_EE_NODEADDR 14
/*
* MII registers registers 0x2000 to 0x3FFF
* There are 32 sets of 32 registers, one set for each possible
* PHY address. Each 32 bit register is split into a 16-bit data
* port and a couple of status bits.
*/
#define SF_MIIADDR_BASE 0x2000
#define SF_MIIADDR_MAX 0x3FFF
#define SF_MII_BLOCKS 32
#define SF_MII_DATAVALID 0x80000000
#define SF_MII_BUSY 0x40000000
#define SF_MII_DATAPORT 0x0000FFFF
#define 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
(SF_MIIADDR_BASE + ((phy) * SF_MII_BLOCKS * sizeof(uint32_t)) + \
((reg) * sizeof(uint32_t)))
/*
* Ethernet extra registers 0x4000 to 0x4FFF
*/
#define SF_TESTMODE 0x4000
#define SF_RX_FRAMEPROC_CTL 0x4004
#define SF_TX_FRAMEPROC_CTL 0x4008
/*
* MAC registers 0x5000 to 0x5FFF
*/
#define SF_MACCFG_1 0x5000
#define SF_MACCFG_2 0x5004
#define SF_BKTOBKIPG 0x5008
#define SF_NONBKTOBKIPG 0x500C
#define SF_COLRETRY 0x5010
#define SF_MAXLEN 0x5014
#define SF_TXNIBBLECNT 0x5018
#define SF_TXBYTECNT 0x501C
#define SF_RETXCNT 0x5020
#define SF_RANDNUM 0x5024
#define SF_RANDNUM_MASK 0x5028
#define SF_TOTALTXCNT 0x5034
#define SF_RXBYTECNT 0x5040
#define SF_TXPAUSETIMER 0x5060
#define SF_VLANTYPE 0x5064
#define SF_MIISTATUS 0x5070
#define SF_MACCFG1_HUGEFRAMES 0x00000001
#define SF_MACCFG1_FULLDUPLEX 0x00000002
#define SF_MACCFG1_AUTOPAD 0x00000004
#define SF_MACCFG1_HDJAM 0x00000008
#define SF_MACCFG1_DELAYCRC 0x00000010
#define SF_MACCFG1_NOBACKOFF 0x00000020
#define SF_MACCFG1_LENGTHCHECK 0x00000040
#define SF_MACCFG1_PUREPREAMBLE 0x00000080
#define SF_MACCFG1_PASSALLRX 0x00000100
#define SF_MACCFG1_PREAM_DETCNT 0x00000200
#define SF_MACCFG1_RX_FLOWENB 0x00000400
#define SF_MACCFG1_TX_FLOWENB 0x00000800
#define SF_MACCFG1_TESTMODE 0x00003000
#define SF_MACCFG1_MIILOOPBK 0x00004000
#define SF_MACCFG1_SOFTRESET 0x00008000
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_MACCFG2_AUTOVLANPAD 0x00000020
/*
* There are the recommended IPG nibble counter settings
* specified in the Adaptec manual for full duplex and
* half duplex operation.
*/
#define SF_IPGT_FDX 0x15
#define SF_IPGT_HDX 0x11
/*
* RX filter registers 0x6000 to 0x6FFF
*/
#define SF_RXFILT_PERFECT_BASE 0x6000
#define SF_RXFILT_PERFECT_MAX 0x60FF
#define SF_RXFILT_PERFECT_SKIP 0x0010
#define SF_RXFILT_PERFECT_CNT 0x0010
#define SF_RXFILT_HASH_BASE 0x6100
#define SF_RXFILT_HASH_MAX 0x62FF
#define SF_RXFILT_HASH_SKIP 0x0010
#define SF_RXFILT_HASH_CNT 0x001F
#define SF_RXFILT_HASH_ADDROFF 0x0000
#define SF_RXFILT_HASH_PRIOOFF 0x0004
#define SF_RXFILT_HASH_VLANOFF 0x0008
/*
* Statistics registers 0x7000 to 0x7FFF
*/
#define SF_STATS_BASE 0x7000
#define SF_STATS_END 0x7FFF
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_STATS_TX_FRAMES 0x0000
#define SF_STATS_TX_SINGLE_COL 0x0004
#define SF_STATS_TX_MULTI_COL 0x0008
#define SF_STATS_TX_CRC_ERRS 0x000C
#define SF_STATS_TX_BYTES 0x0010
#define SF_STATS_TX_DEFERRED 0x0014
#define SF_STATS_TX_LATE_COL 0x0018
#define SF_STATS_TX_PAUSE 0x001C
#define SF_STATS_TX_CTL_FRAME 0x0020
#define SF_STATS_TX_EXCESS_COL 0x0024
#define SF_STATS_TX_EXCESS_DEF 0x0028
#define SF_STATS_TX_MULTI 0x002C
#define SF_STATS_TX_BCAST 0x0030
#define SF_STATS_TX_FRAME_LOST 0x0034
#define SF_STATS_RX_FRAMES 0x0038
#define SF_STATS_RX_CRC_ERRS 0x003C
#define SF_STATS_RX_ALIGN_ERRS 0x0040
#define SF_STATS_RX_BYTES 0x0044
#define SF_STATS_RX_PAUSE 0x0048
#define SF_STATS_RX_CTL_FRAME 0x004C
#define SF_STATS_RX_UNSUP_FRAME 0x0050
#define SF_STATS_RX_GIANTS 0x0054
#define SF_STATS_RX_RUNTS 0x0058
#define SF_STATS_RX_JABBER 0x005C
#define SF_STATS_RX_FRAGMENTS 0x0060
#define SF_STATS_RX_64 0x0064
#define SF_STATS_RX_65_127 0x0068
#define SF_STATS_RX_128_255 0x006C
#define SF_STATS_RX_256_511 0x0070
#define SF_STATS_RX_512_1023 0x0074
#define SF_STATS_RX_1024_1518 0x0078
#define SF_STATS_RX_FRAME_LOST 0x007C
#define SF_STATS_TX_UNDERRUN 0x0080
/*
* TX frame processor instruction space 0x8000 to 0x9FFF
*/
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_TXGFP_MEM_BASE 0x8000
#define SF_TXGFP_MEM_END 0x8FFF
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Number of bytes of an GFP instruction. */
#define SF_GFP_INST_BYTES 6
/*
* RX frame processor instruction space 0xA000 to 0xBFFF
*/
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_RXGFP_MEM_BASE 0xA000
#define SF_RXGFP_MEM_END 0xBFFF
/*
* Ethernet FIFO access space 0xC000 to 0xDFFF
*/
/*
* Reserved 0xE000 to 0xFFFF
*/
/*
* Descriptor data structures.
*/
/*
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
* RX buffer descriptor type 0, 32-bit addressing.
*/
struct sf_rx_bufdesc_type0 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_addrlo;
#define SF_RX_DESC_VALID 0x00000001
#define SF_RX_DESC_END 0x00000002
};
/*
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
* RX buffer descriptor type 1, 64-bit addressing.
*/
struct sf_rx_bufdesc_type1 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint64_t sf_addr;
};
/*
* RX completion descriptor, type 0 (short).
*/
struct sf_rx_cmpdesc_type0 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_rx_status1;
#define SF_RX_CMPDESC_LEN 0x0000ffff
#define SF_RX_CMPDESC_EIDX 0x07ff0000
#define SF_RX_CMPDESC_STAT1 0x38000000
#define SF_RX_CMPDESC_ID 0x40000000
};
/*
* RX completion descriptor, type 1 (basic). Includes vlan ID
* if this is a vlan-addressed packet, plus extended status.
*/
struct sf_rx_cmpdesc_type1 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_rx_status1;
uint32_t sf_rx_status2;
#define SF_RX_CMPDESC_VLAN 0x0000ffff
#define SF_RX_CMPDESC_STAT2 0xffff0000
};
/*
* RX completion descriptor, type 2 (checksum). Includes partial TCP/IP
* checksum instead of vlan tag, plus extended status.
*/
struct sf_rx_cmpdesc_type2 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_rx_status1;
uint32_t sf_rx_status2;
#define SF_RX_CMPDESC_CSUM2 0x0000ffff
};
/*
* RX completion descriptor type 3 (full). Includes timestamp, partial
* TCP/IP checksum, vlan tag plus priority, two extended status fields.
*/
struct sf_rx_cmpdesc_type3 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_rx_status1;
uint32_t sf_rx_status2;
uint32_t sf_rx_status3;
#define SF_RX_CMPDESC_CSUM3 0xffff0000
#define SF_RX_CMPDESC_VLANPRI 0x0000ffff
uint32_t sf_rx_timestamp;
};
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_RXSTAT1_QUEUE 0x08000000
#define SF_RXSTAT1_FIFOFULL 0x10000000
#define SF_RXSTAT1_OK 0x20000000
#define SF_RXSTAT2_FRAMETYPE_MASK 0x00070000
#define SF_RXSTAT2_FRAMETYPE_UNKN 0x00000000
#define SF_RXSTAT2_FRAMETYPE_IPV4 0x00010000
#define SF_RXSTAT2_FRAMETYPE_IPV6 0x00020000
#define SF_RXSTAT2_FRAMETYPE_IPX 0x00030000
#define SF_RXSTAT2_FRAMETYPE_ICMP 0x00040000
#define SF_RXSTAT2_FRAMETYPE_UNSPRT 0x00050000
#define SF_RXSTAT2_UDP 0x00080000
#define SF_RXSTAT2_TCP 0x00100000
#define SF_RXSTAT2_FRAG 0x00200000
#define SF_RXSTAT2_PCSUM_OK 0x00400000 /* partial checksum ok */
#define SF_RXSTAT2_CSUM_BAD 0x00800000 /* TCP/IP checksum bad */
#define SF_RXSTAT2_CSUM_OK 0x01000000 /* TCP/IP checksum ok */
#define SF_RXSTAT2_VLAN 0x02000000
#define SF_RXSTAT2_BADRXCODE 0x04000000
#define SF_RXSTAT2_DRIBBLE 0x08000000
#define SF_RXSTAT2_ISL_CRCERR 0x10000000
#define SF_RXSTAT2_CRCERR 0x20000000
#define SF_RXSTAT2_HASH 0x40000000
#define SF_RXSTAT2_PERFECT 0x80000000
#define SF_RXSTAT2_MASK 0xFFFF0000
#define SF_RXSTAT3_ISL 0x00008000
#define SF_RXSTAT3_PAUSE 0x00004000
#define SF_RXSTAT3_CONTROL 0x00002000
#define SF_RXSTAT3_HEADER 0x00001000
#define SF_RXSTAT3_TRAILER 0x00000800
#define SF_RXSTAT3_START_IDX_MASK 0x000007FF
struct sf_frag {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_addr;
uint16_t sf_fraglen;
uint16_t sf_pktlen;
};
struct sf_frag_msdos {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_pktlen;
uint16_t sf_fraglen;
uint32_t sf_addr;
};
/*
* TX frame descriptor type 0, 32-bit addressing. One descriptor can
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
* be used to map multiple packet fragments. Note that the number of
* fragments can be variable depending on how the descriptor spacing
* is specified in the TX descriptor queue control register.
* We always use a spacing of 128 bytes, and a skipfield length of 8
* bytes: this means 16 bytes for the descriptor, including the skipfield,
* with 121 bytes left for fragment maps. Each fragment requires 8 bytes,
* which allows for 14 fragments per descriptor. The total size of the
* transmit buffer queue is limited to 16384 bytes, so with a spacing of
* 128 bytes per descriptor, we have room for 128 descriptors in the queue.
*/
struct sf_tx_bufdesc_type0 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_tx_ctrl;
#define SF_TX_DESC_CRCEN 0x01000000
#define SF_TX_DESC_CALTCP 0x02000000
#define SF_TX_DESC_END 0x04000000
#define SF_TX_DESC_INTR 0x08000000
#define SF_TX_DESC_ID 0xb0000000
uint32_t sf_tx_frag;
/*
* Depending on descriptor spacing/skip field length it
* can have fixed number of struct sf_frag.
* struct sf_frag sf_frags[14];
*/
};
/*
* TX buffer descriptor type 1, 32-bit addressing. Each descriptor
* maps a single fragment.
*/
struct sf_tx_bufdesc_type1 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_tx_ctrl;
#define SF_TX_DESC_FRAGLEN 0x0000ffff
#define SF_TX_DESC_FRAGCNT 0x00ff0000
uint32_t sf_addrlo;
};
/*
* TX buffer descriptor type 2, 64-bit addressing. Each descriptor
* maps a single fragment.
*/
struct sf_tx_bufdesc_type2 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_tx_ctrl;
uint32_t sf_tx_reserved;
uint64_t sf_addr;
};
/* TX buffer descriptor type 3 is not defined. */
/*
* TX frame descriptor type 4, 32-bit addressing. This is a special
* case of the type 0 descriptor, identical except that the fragment
* address and length fields are ordered differently. This is done
* to optimize copies in MS-DOS and OS/2 drivers.
*/
struct sf_tx_bufdesc_type4 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_tx_ctrl;
uint32_t sf_tx_frag;
/*
* Depending on descriptor spacing/skip field length it
* can have fixed number of struct sf_frag_msdos.
*
* struct sf_frag_msdos sf_frags[14];
*/
};
/*
* Transmit completion queue descriptor formats.
*/
/*
* Transmit DMA completion descriptor, type 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
#define SF_TXCMPTYPE_DMA 0x80000000
#define SF_TXCMPTYPE_TX 0xa0000000
struct sf_tx_cmpdesc_type0 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_tx_status1;
#define SF_TX_CMPDESC_IDX 0x00007fff
#define SF_TX_CMPDESC_HIPRI 0x00008000
#define SF_TX_CMPDESC_STAT 0x1fff0000
#define SF_TX_CMPDESC_TYPE 0xe0000000
};
/*
* Transmit completion descriptor, type 1.
*/
struct sf_tx_cmpdesc_type1 {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_tx_status1;
uint32_t sf_tx_status2;
};
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_TXSTAT_CRCERR 0x00010000
#define SF_TXSTAT_LENCHECKERR 0x00020000
#define SF_TXSTAT_LENRANGEERR 0x00040000
#define SF_TXSTAT_TX_OK 0x00080000
#define SF_TXSTAT_TX_DEFERED 0x00100000
#define SF_TXSTAT_EXCESS_DEFER 0x00200000
#define SF_TXSTAT_EXCESS_COLL 0x00400000
#define SF_TXSTAT_LATE_COLL 0x00800000
#define SF_TXSTAT_TOOBIG 0x01000000
#define SF_TXSTAT_TX_UNDERRUN 0x02000000
#define SF_TXSTAT_CTLFRAME_OK 0x04000000
#define SF_TXSTAT_PAUSEFRAME_OK 0x08000000
#define SF_TXSTAT_PAUSED 0x10000000
/* Statistics counters. */
struct sf_stats {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
uint64_t sf_tx_frames;
uint32_t sf_tx_single_colls;
uint32_t sf_tx_multi_colls;
uint32_t sf_tx_crcerrs;
uint64_t sf_tx_bytes;
uint32_t sf_tx_deferred;
uint32_t sf_tx_late_colls;
uint32_t sf_tx_pause_frames;
uint32_t sf_tx_control_frames;
uint32_t sf_tx_excess_colls;
uint32_t sf_tx_excess_defer;
uint32_t sf_tx_mcast_frames;
uint32_t sf_tx_bcast_frames;
uint32_t sf_tx_frames_lost;
uint64_t sf_rx_frames;
uint32_t sf_rx_crcerrs;
uint32_t sf_rx_alignerrs;
uint64_t sf_rx_bytes;
uint32_t sf_rx_pause_frames;
uint32_t sf_rx_control_frames;
uint32_t sf_rx_unsup_control_frames;
uint32_t sf_rx_giants;
uint32_t sf_rx_runts;
uint32_t sf_rx_jabbererrs;
uint32_t sf_rx_fragments;
uint64_t sf_rx_pkts_64;
uint64_t sf_rx_pkts_65_127;
uint64_t sf_rx_pkts_128_255;
uint64_t sf_rx_pkts_256_511;
uint64_t sf_rx_pkts_512_1023;
uint64_t sf_rx_pkts_1024_1518;
uint32_t sf_rx_frames_lost;
uint32_t sf_tx_underruns;
uint32_t sf_tx_gfp_stall;
uint32_t sf_rx_gfp_stall;
};
/*
* register space access macros
*/
#define CSR_WRITE_4(sc, reg, val) \
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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_write_4((sc)->sf_res, reg, val)
#define CSR_READ_4(sc, 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
bus_read_4((sc)->sf_res, reg)
#define CSR_READ_1(sc, 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
bus_read_1((sc)->sf_res, reg)
struct sf_type {
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_vid;
uint16_t sf_did;
char *sf_name;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_sdid;
char *sf_sname;
};
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
/* Use Tx descriptor type 2 : 64bit buffer descriptor */
#define sf_tx_rdesc sf_tx_bufdesc_type2
/* Use Rx descriptor type 1 : 64bit buffer descriptor */
#define sf_rx_rdesc sf_rx_bufdesc_type1
/* Use Tx completion type 0 */
#define sf_tx_rcdesc sf_tx_cmpdesc_type0
/* Use Rx completion type 2 : checksum */
#define sf_rx_rcdesc sf_rx_cmpdesc_type2
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. sf(4) doesn't use bit fileds anymore to support endianness. Tested by: bruffer (initial version)
2008-01-21 06:38:23 +00:00
#define SF_TX_DLIST_CNT 256
#define SF_RX_DLIST_CNT 256
#define SF_TX_CLIST_CNT 1024
#define SF_RX_CLIST_CNT 1024
#define SF_TX_DLIST_SIZE (sizeof(struct sf_tx_rdesc) * SF_TX_DLIST_CNT)
#define SF_TX_CLIST_SIZE (sizeof(struct sf_tx_rcdesc) * SF_TX_CLIST_CNT)
#define SF_RX_DLIST_SIZE (sizeof(struct sf_rx_rdesc) * SF_RX_DLIST_CNT)
#define SF_RX_CLIST_SIZE (sizeof(struct sf_rx_rcdesc) * SF_RX_CLIST_CNT)
#define SF_RING_ALIGN 256
#define SF_RX_ALIGN sizeof(uint32_t)
#define SF_MAXTXSEGS 16
#define SF_ADDR_LO(x) ((uint64_t)(x) & 0xffffffff)
#define SF_ADDR_HI(x) ((uint64_t)(x) >> 32)
#define SF_TX_DLIST_ADDR(sc, i) \
((sc)->sf_rdata.sf_tx_ring_paddr + sizeof(struct sf_tx_rdesc) * (i))
#define SF_TX_CLIST_ADDR(sc, i) \
((sc)->sf_rdata.sf_tx_cring_paddr + sizeof(struct sf_tx_crdesc) * (i))
#define SF_RX_DLIST_ADDR(sc, i) \
((sc)->sf_rdata.sf_rx_ring_paddr + sizeof(struct sf_rx_rdesc) * (i))
#define SF_RX_CLIST_ADDR(sc, i) \
((sc)->sf_rdata.sf_rx_cring_paddr + sizeof(struct sf_rx_rcdesc) * (i))
#define SF_INC(x, y) (x) = ((x) + 1) % y
#define SF_MAX_FRAMELEN 1536
#define SF_TX_THRESHOLD_UNIT 16
#define SF_MAX_TX_THRESHOLD (SF_MAX_FRAMELEN / SF_TX_THRESHOLD_UNIT)
#define SF_MIN_TX_THRESHOLD (128 / SF_TX_THRESHOLD_UNIT)
struct sf_txdesc {
struct mbuf *tx_m;
int ndesc;
bus_dmamap_t tx_dmamap;
};
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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_rxdesc {
struct mbuf *rx_m;
bus_dmamap_t rx_dmamap;
};
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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_chain_data {
bus_dma_tag_t sf_parent_tag;
bus_dma_tag_t sf_tx_tag;
struct sf_txdesc sf_txdesc[SF_TX_DLIST_CNT];
bus_dma_tag_t sf_rx_tag;
struct sf_rxdesc sf_rxdesc[SF_RX_DLIST_CNT];
bus_dma_tag_t sf_tx_ring_tag;
bus_dma_tag_t sf_rx_ring_tag;
bus_dma_tag_t sf_tx_cring_tag;
bus_dma_tag_t sf_rx_cring_tag;
bus_dmamap_t sf_tx_ring_map;
bus_dmamap_t sf_rx_ring_map;
bus_dmamap_t sf_rx_sparemap;
bus_dmamap_t sf_tx_cring_map;
bus_dmamap_t sf_rx_cring_map;
int sf_tx_prod;
int sf_tx_cnt;
int sf_txc_cons;
int sf_rxc_cons;
};
struct sf_ring_data {
struct sf_tx_rdesc *sf_tx_ring;
bus_addr_t sf_tx_ring_paddr;
struct sf_tx_rcdesc *sf_tx_cring;
bus_addr_t sf_tx_cring_paddr;
struct sf_rx_rdesc *sf_rx_ring;
bus_addr_t sf_rx_ring_paddr;
struct sf_rx_rcdesc *sf_rx_cring;
bus_addr_t sf_rx_cring_paddr;
};
struct sf_softc {
struct ifnet *sf_ifp; /* interface info */
2006-09-15 11:01:23 +00:00
device_t sf_dev; /* device info */
void *sf_intrhand; /* interrupt handler cookie */
struct resource *sf_irq; /* irq resource descriptor */
struct resource *sf_res; /* mem/ioport resource */
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 sf_restype;
int sf_rid;
struct sf_type *sf_info; /* Starfire adapter info */
device_t 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
struct sf_chain_data sf_cdata;
struct sf_ring_data sf_rdata;
int sf_if_flags;
Overhaul sf(4) to make it run on all architectures and implement checksum offoload by downloading AIC-6915 firmware. Changes are o Header file cleanup. o Simplified probe logic. o s/u_int{8,16,32}_t/uint{8,16,32}_t/g o K&R -> ANSI C. o In register access function, added support both memory mapped and IO space register acccess. The function will dynamically detect which method would be choosed. o sf_setperf() was modified to support strict-alignment architectures. o Use SF_MII_DATAPORT instead of hardcoded value 0xffff. o Added link state/speed, duplex changes handling task q. The task q is also responsible for flow control settings. o Always hornor link up/down state reported by mii layers. The link state information is used in sf_start() to determine whether we got a valid link. o Added experimental flow-control setup. It was commented out but will be activated once we have flow-cotrol infrastructure in mii layer. o Simplify IFF_UP/IFCAP_POLLING and IFF_PROMISC handling logic. Rx filter always honors promiscuous mode. o Implemented suspend/resume methods. o Reorganized Rx filter routine so promiscuous mode changes doesn't require interface re-initialization. o Reimplemnted driver probe routine such that it looks for matching device from supported hardware list table. This change will help to add newer hardware revision to the driver. o Use ETHER_ADDR_LEN instead of hardcoded value. o Prefer memory space register mapping over I/O space as the hardware requires lots of register access to get various consumer/producer index. Failing to get memory space mapping, sf(4) falls back to I/O space mapping. Use of memory space register mapping requires somewhat large memory space(512K), though. o Switch to simpler bus_{read,write}_{1,2,4}. o Use PCIR_BAR macro to get BARs. o Program PCI cache line size if the cache line size was set to 0 and enable PCI MWI. o Add a new sysctl node 'dev.sf.N.stats' that shows various MAC counters for Rx/Tx statistics. o Add a sysctl node to configure interrupt moderation timer. The timer defers interrupts generation until time specified in timer control register is expired. The value in the timer register is in units of 102.4us. The allowable range for the timer is 0 - 31 (0 ~ 3.276ms). The default value is 1(102.4us). Users can change the timer value with dev.sf.N.int_mod sysctl(8) variable/loader(8) tunable. o bus_dma(9) conversion - Enable 64bit DMA addressing. - Enable 64bit descriptor format support. - Apply descriptor ring alignment requirements(256 bytes alignment). - Apply Rx buffer address alignment requirements(4 bytes alignment). - Apply 4GB boundary restrictions(Tx/Rx ring and its completion ring should live in the same 4GB address space.) - Set number of allowable number of DMA segments to 16. In fact, AIC-6915 doesn't have a limit for number of DMA segments but it would be waste of Tx descriptor resource if we allow more than 16. - Rx/Tx side bus_dmamap_load_mbuf_sg(9) support. - Added alignment fixup code for strict-alignment architectures. - Added endianness support code in Tx/Rx descriptor access. With these changes sf(4) should work on all platforms. o Don't set if_mtu in device attach, it's handled in ether_ifattach. o Use our own callout to drive watchdog timer. o Enable VLAN oversized frames and announce sf(4)'s VLAN capability to upper layer. o In sf_detach(), remove mtx_initialized KASSERT as it's not possible to get there without initialzing the mutex. Also mark that we're about to detaching so active bpf listeners do not panic the system. o To reduce PCI register access cycles, Rx completion ring is directly scanned instead of reading consumer/producer index registers. In theory, Tx completion ring also can be directly scanned. However the completion ring is composed of two types completion(1 for Tx done and 1 and DMA done). So reading producer index via register access would be more safer way to detect the ring wrap-around. o In sf_rxeof(), don't use m_devget(9) to align recevied frames. The alignment is required only for strict-alignment architectures and now the alignment is handled by sf_fixup_rx() if required. The removal of the copy operation in fast path should increase Rx performance a lot on non-strict-alignemnt architectures such as i386 and amd64. o In sf_newbuf(), don't set descriptor valid bit as sf(4) is programmed to run with normal mode. In normal mode, the valid bit have no meaning. The valid bit should be used only when the hardware uses polling(prefetch) mode. The end of descriptor queue bit could be used if needed, but sf(4) relys on auto-wrapping of hardware on 256 descriptor queue entries so both valid and descriptor end bit are not used anymore. o Don't disable generation of Tx DMA completion as said in datasheet and use the Tx DMA completion entry instead of relying on Tx done completion entry. Also added additional Tx completion entry type check in Tx completion handler. o Don't blindly reset watchdog timer in sf_txeof(). sf(4) now unarm the the watchdog only if there are no active Tx descriptors in Tx queue. o Don't manually update various counters in driver, instead, use built-in MAC statistic registers to update them. The statistic registers are updated in every second. o Modified Tx underrun handlers to increase the threshold value in units of 256 bytes. Previously it used to increase 16 bytes at a time which seems to take too long to stabalize whenever Tx underrun occurrs. o In interrupt handler, additional check for the interrupt is performed such that interrupts only for this device is allowed to process descriptor rings. Because reading SF_ISR register clears all interrtups, nuke writing to a SF_ISR register. o Tx underrun is abonormal condition and SF_ISR_ABNORMALINTR includes the interrupt. So there is no need to inspect the Tx underrun again in main interrupt loop. o Don't blindly reinitialize hardware for abnormal interrupt condition. sf(4) reintializes the hardware only when it encounters DMA error which requires an explicit hardware reinitialization. o Fix a long standing bug that incorrectly clears MAC statistic registers in sf_init_locked. o Added strict-alignment safe way of ethernet address reprogramming as IF_LLADDR may return unaligned address. o Move sf_reset() to sf_init_locked in order to always reset the hardware to a known state prior to configuring hardware. o Set default Rx DMA, Tx DMA paramters as shown in datasheet. o Enable PCI busmaster logic and autopadding for VLAN frames. o Rework sf_encap. - Previously sf(4) used to type 0 of Tx descriptor with padding enabled to store driver private data. Emebedding private data structures into descriptors is bad idea as the structure size would be different between 64bit and 32bit architectures. The type 0 descriptor allows fixed number of DMA segments in a descriptor format and provides relatively simple interface to manage multi-fragmented frames. However, it wastes lots of Tx descriptors as not all frames are fragmented as the number of allowable segments in a descriptor. - To overcome the limitation of type 0 descriptor, switch to type 2 descriptor which allows 64bit DMA addressing and can handle unliumited number of fragmented DMA segments. The drawback of type 2 descriptor is in its complexity in managing descriptors as driver should handle the end of Tx ring manually. - Manually set Tx desciptor queue end mark and record number of used descriptors to reclaim used descriptors in sf_txeof(). o Rework sf_start. - Honor link up/down state before attempting transmission. - Because sf(4) uses only one of two Tx queues, use low priority queue instead of high one. This will remove one shift operation in each Tx kick command. - Cache last produder index into softc such that subsequenet Tx operation doesn't need to access producer index register. o Rewrote sf_stats_update to include all available MAC statistic counters. o Employ AIC-6915 firmware from Adaptec and implement firmware download routine and TCP/UDP checksum offload. Partial checksum offload support was commented out due to the possibility of firmware bug in RxGFP. The firmware can strip VLAN tag in Rx path but the lack of firmware assistance of VLAN tag insertion in transmit side made it useless on FreeBSD. Unlike checksum offload, FreeBSD requires both Tx/Rx hardware VLAN assistance capability. The firmware may also detect wakeup frame and can wake system up from states other than D0. However, the lack of wakeup support form D3cold state keep me from adding WOL capability. Also detecting WOL frame requires firmware support but it's not yet known to me whether the firmware can process the WOL frame. o Changed *_ADDR_HIADDR to *_ADDR_HI to match other definitions of registers. o Added definitioan to interrupt moderation related constants. o Redefined SF_INTRS to include Tx DMA done and DMA errors. Removed Tx done as it's not needed anymore. o Added definition for Rx/Tx DMA high priority threshold. o Nuked unused marco SF_IDX_LO, SF_IDX_HI. o Added complete MAC statistic register definition. o Modified sf_stats structure to hold all MAC statistic regiters. o Nuke various driver private padding data in Tx/Rx descriptor definition. sf(4) no longer requires private padding. Also remove unused padding related definitions. This greatly simplifies descriptor manipulation on 64bit architectures. o Becase we no longer pad driver private data into descriptor, remove deprecated/not-applicable comments for padding. o Redefine Rx/Tx desciptor status. 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 callout sf_co;
int sf_watchdog_timer;
struct task sf_link_task;
int sf_link;
int sf_suspended;
int sf_detach;
uint32_t sf_txthresh;
int sf_int_mod;
struct sf_stats sf_statistics;
struct mtx sf_mtx;
#ifdef DEVICE_POLLING
int rxcycles;
2004-11-17 23:33:56 +00:00
#endif /* DEVICE_POLLING */
};
Change and clean the mutex lock interface. mtx_enter(lock, type) becomes: mtx_lock(lock) for sleep locks (MTX_DEF-initialized locks) mtx_lock_spin(lock) for spin locks (MTX_SPIN-initialized) similarily, for releasing a lock, we now have: mtx_unlock(lock) for MTX_DEF and mtx_unlock_spin(lock) for MTX_SPIN. We change the caller interface for the two different types of locks because the semantics are entirely different for each case, and this makes it explicitly clear and, at the same time, it rids us of the extra `type' argument. The enter->lock and exit->unlock change has been made with the idea that we're "locking data" and not "entering locked code" in mind. Further, remove all additional "flags" previously passed to the lock acquire/release routines with the exception of two: MTX_QUIET and MTX_NOSWITCH The functionality of these flags is preserved and they can be passed to the lock/unlock routines by calling the corresponding wrappers: mtx_{lock, unlock}_flags(lock, flag(s)) and mtx_{lock, unlock}_spin_flags(lock, flag(s)) for MTX_DEF and MTX_SPIN locks, respectively. Re-inline some lock acq/rel code; in the sleep lock case, we only inline the _obtain_lock()s in order to ensure that the inlined code fits into a cache line. In the spin lock case, we inline recursion and actually only perform a function call if we need to spin. This change has been made with the idea that we generally tend to avoid spin locks and that also the spin locks that we do have and are heavily used (i.e. sched_lock) do recurse, and therefore in an effort to reduce function call overhead for some architectures (such as alpha), we inline recursion for this case. Create a new malloc type for the witness code and retire from using the M_DEV type. The new type is called M_WITNESS and is only declared if WITNESS is enabled. Begin cleaning up some machdep/mutex.h code - specifically updated the "optimized" inlined code in alpha/mutex.h and wrote MTX_LOCK_SPIN and MTX_UNLOCK_SPIN asm macros for the i386/mutex.h as we presently need those. Finally, caught up to the interface changes in all sys code. Contributors: jake, jhb, jasone (in no particular order)
2001-02-09 06:11:45 +00:00
#define SF_LOCK(_sc) mtx_lock(&(_sc)->sf_mtx)
#define SF_UNLOCK(_sc) mtx_unlock(&(_sc)->sf_mtx)
#define SF_LOCK_ASSERT(_sc) mtx_assert(&(_sc)->sf_mtx, MA_OWNED)
#define SF_TIMEOUT 1000