d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
9b3a48ee6f
freebsd-dev/sys/dev/ti/if_ti.c

4059 lines
109 KiB
C
Raw Normal View History

/* -> /*- for license, minor formatting changes
2005-01-07 02:29:27 +00:00
/*-
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
* Copyright (c) 1997, 1998, 1999
* Bill Paul <wpaul@ctr.columbia.edu>. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Bill Paul.
* 4. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Alteon Networks Tigon PCI gigabit ethernet driver for FreeBSD.
* Manuals, sample driver and firmware source kits are available
* from http://www.alteon.com/support/openkits.
Whitespace nits
2004-07-05 22:35:18 +00:00
*
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
* Written by Bill Paul <wpaul@ctr.columbia.edu>
* Electrical Engineering Department
* Columbia University, New York City
*/
/*
* The Alteon Networks Tigon chip contains an embedded R4000 CPU,
* gigabit MAC, dual DMA channels and a PCI interface unit. NICs
* using the Tigon may have anywhere from 512K to 2MB of SRAM. The
* Tigon supports hardware IP, TCP and UCP checksumming, multicast
* filtering and jumbo (9014 byte) frames. The hardware is largely
* controlled by firmware, which must be loaded into the NIC during
* initialization.
*
* The Tigon 2 contains 2 R4000 CPUs and requires a newer firmware
* revision, which supports new features such as extended commands,
* extended jumbo receive ring desciptors and a mini receive ring.
*
* Alteon Networks is to be commended for releasing such a vast amount
* of development material for the Tigon NIC without requiring an NDA
* (although they really should have done it a long time ago). With
* any luck, the other vendors will finally wise up and follow Alteon's
* stellar example.
*
* The firmware for the Tigon 1 and 2 NICs is compiled directly into
* this driver by #including it as a C header file. This bloats the
* driver somewhat, but it's the easiest method considering that the
* driver code and firmware code need to be kept in sync. The source
* for the firmware is not provided with the FreeBSD distribution since
* compiling it requires a GNU toolchain targeted for mips-sgi-irix5.3.
*
* The following people deserve special thanks:
* - Terry Murphy of 3Com, for providing a 3c985 Tigon 1 board
* for testing
* - Raymond Lee of Netgear, for providing a pair of Netgear
* GA620 Tigon 2 boards for testing
* - Ulf Zimmermann, for bringing the GA260 to my attention and
* convincing me to write this driver.
* - Andrew Gallatin for providing FreeBSD/Alpha support.
*/
Use __FBSDID rather than rcsid[].
2003-04-03 21:36:33 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#include "opt_ti.h"
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sockio.h>
#include <sys/mbuf.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
Add missing <sys/module.h> includes
2004-05-30 20:00:41 +00:00
#include <sys/module.h>
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
#include <sys/socket.h>
#include <sys/queue.h>
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#include <sys/conf.h>
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
#include <sys/sf_buf.h>
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
#include <net/if.h>
#include <net/if_arp.h>
#include <net/ethernet.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <net/if_vlan_var.h>
Make vlan(4) loadable, unloadable, and clonable. As a side effect, interfaces must now always enable VLAN support. Reviewed by: jlemon MFC after: 3 weeks
2001-09-05 21:10:28 +00:00
#include <net/bpf.h>
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
#include <netinet/in_systm.h>
#include <netinet/in.h>
#include <netinet/ip.h>
#include <machine/bus.h>
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
#include <machine/resource.h>
#include <sys/bus.h>
#include <sys/rman.h>
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
#ifdef TI_SF_BUF_JUMBO
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
#include <vm/vm.h>
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#include <vm/vm_page.h>
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
#endif
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Prefer new location of pci include files (which have only been in the tree for two or more years now), except in a few places where there's code to be compatible with older versions of FreeBSD.
2003-08-22 07:20:27 +00:00
#include <dev/pci/pcireg.h>
#include <dev/pci/pcivar.h>
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#include <sys/tiio.h>
The if_ti Tigon I/II driver has moved to /sys/dev/ti
2005-12-10 00:38:33 +00:00
#include <dev/ti/if_tireg.h>
#include <dev/ti/ti_fw.h>
#include <dev/ti/ti_fw2.h>
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Export sysctl node for various interrupt moderation parameters and have administrators control them. ti(4) provides a character device to control various other features of driver via ioctls but users had to write their own code to manipulate these parameters. It seems some default values for these parameters are not optimal on today's system but leave it as it was and let administrators change them. The following parameters could be changed: dev.ti.%d.rx_coal_ticks dev.ti.%d.rx_max_coal_bds dev.ti.%d.tx_coal_ticks dev.ti.%d.tx_max_coal_bds dev.ti.%d.tx_buf_ratio dev.ti.%d.stat_ticks The interface has to be brought down and up again before a change takes effect. ti(4) controller supports hardware MAC counters with additional DMA statistics. So it's doable to export these counters via sysctl interface. Unfortunately, these counters are cumulative such that driver have to either send an explicit clear command to controller after extracting them or have to maintain internal counters to get actual changes. Neither look good to me so counters were not exported via sysctl.
2011-11-14 19:10:20 +00:00
#include <sys/sysctl.h>
Remove TCP/UDP checksum offloading feature for IP fragmented datagrams. Traditionally upper stack fragmented packets without computing TCP/UDP checksum and these datagrams were passed to driver. But there are chances that other packets slip into the interface queue in SMP world. If this happens firmware running on MIPS 4000 processor in the controller would see mixed packets and it shall send out corrupted packets. While I'm here simplify checksum offloading setup. MFC After: 1 week
2012-11-01 05:39:21 +00:00
#define TI_CSUM_FEATURES (CSUM_IP | CSUM_TCP | CSUM_UDP)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* We can only turn on header splitting if we're using extended receive
* BDs.
*/
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
#if defined(TI_JUMBO_HDRSPLIT) && !defined(TI_SF_BUF_JUMBO)
#error "options TI_JUMBO_HDRSPLIT requires TI_SF_BUF_JUMBO"
#endif /* TI_JUMBO_HDRSPLIT && !TI_SF_BUF_JUMBO */
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
typedef enum {
TI_SWAP_HTON,
TI_SWAP_NTOH
} ti_swap_type;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Various supported device vendors/types and their names.
*/
Remove duplicate const specifiers in many drivers (I hope I got all of them, please let me know if not). Most of these are of the form: static const struct bzzt_type { [...list of members...] } const bzzt_devs[] = { [...list of initializers...] }; The second const is unnecessary, as arrays cannot be modified anyway, and if the elements are const, the whole thing is const automatically (e.g. it is placed in .rodata). I have verified this does not change the binary output of a full kernel build (except for build timestamps embedded in the object files). Reviewed by: yongari, marius MFC after: 1 week
2012-11-05 19:16:27 +00:00
static const struct ti_type ti_devs[] = {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{ ALT_VENDORID, ALT_DEVICEID_ACENIC,
Tweak probe message so that 1000baseSX and 1000baseT cards are explicitly identified.
2000-07-20 23:43:45 +00:00
"Alteon AceNIC 1000baseSX Gigabit Ethernet" },
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
{ ALT_VENDORID, ALT_DEVICEID_ACENIC_COPPER,
Tweak probe message so that 1000baseSX and 1000baseT cards are explicitly identified.
2000-07-20 23:43:45 +00:00
"Alteon AceNIC 1000baseT Gigabit Ethernet" },
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{ TC_VENDORID, TC_DEVICEID_3C985,
"3Com 3c985-SX Gigabit Ethernet" },
{ NG_VENDORID, NG_DEVICEID_GA620,
Add support for the Netgear GA620T copper gigabit card.
2000-08-02 18:49:17 +00:00
"Netgear GA620 1000baseSX Gigabit Ethernet" },
{ NG_VENDORID, NG_DEVICEID_GA620T,
"Netgear GA620 1000baseT Gigabit Ethernet" },
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{ SGI_VENDORID, SGI_DEVICEID_TIGON,
"Silicon Graphics Gigabit Ethernet" },
Add the vendor/device ID for the Farallon PN9000SX gigabit ethernet card, which is apparently also a Tigon 2 device.
2000-01-18 00:26:29 +00:00
{ DEC_VENDORID, DEC_DEVICEID_FARALLON_PN9000SX,
"Farallon PN9000SX Gigabit Ethernet" },
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{ 0, 0, NULL }
};
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
static d_open_t ti_open;
static d_close_t ti_close;
static d_ioctl_t ti_ioctl2;
static struct cdevsw ti_cdevsw = {
Device megapatch 4/6: Introduce d_version field in struct cdevsw, this must always be initialized to D_VERSION. Flip sense of D_NOGIANT flag to D_NEEDGIANT, this involves removing four D_NOGIANT flags and adding 145 D_NEEDGIANT flags.
2004-02-21 21:10:55 +00:00
.d_version = D_VERSION,
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
.d_flags = 0,
Gigacommit to improve device-driver source compatibility between branches: Initialize struct cdevsw using C99 sparse initializtion and remove all initializations to default values. This patch is automatically generated and has been tested by compiling LINT with all the fields in struct cdevsw in reverse order on alpha, sparc64 and i386. Approved by: re(scottl)
2003-03-03 12:15:54 +00:00
.d_open = ti_open,
.d_close = ti_close,
.d_ioctl = ti_ioctl2,
.d_name = "ti",
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
};
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int ti_probe(device_t);
static int ti_attach(device_t);
static int ti_detach(device_t);
static void ti_txeof(struct ti_softc *);
static void ti_rxeof(struct ti_softc *);
static void ti_stats_update(struct ti_softc *);
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
static int ti_encap(struct ti_softc *, struct mbuf **);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void ti_intr(void *);
static void ti_start(struct ifnet *);
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
static void ti_start_locked(struct ifnet *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int ti_ioctl(struct ifnet *, u_long, caddr_t);
static void ti_init(void *);
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
static void ti_init_locked(void *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void ti_init2(struct ti_softc *);
static void ti_stop(struct ti_softc *);
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
static void ti_watchdog(void *);
Fix function prototype for device_shutdown method.
2007-11-22 02:45:00 +00:00
static int ti_shutdown(device_t);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int ti_ifmedia_upd(struct ifnet *);
Introduce ti_ifmedia_upd_locked() to use in driver initialization and add missing driver lock for both ti_ifmedia_upd() and ti_ifmedia_sts().
2011-11-04 21:30:46 +00:00
static int ti_ifmedia_upd_locked(struct ti_softc *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static void ti_ifmedia_sts(struct ifnet *, struct ifmediareq *);
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
static uint32_t ti_eeprom_putbyte(struct ti_softc *, int);
static uint8_t ti_eeprom_getbyte(struct ti_softc *, int, uint8_t *);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int ti_read_eeprom(struct ti_softc *, caddr_t, int, int);
static void ti_add_mcast(struct ti_softc *, struct ether_addr *);
static void ti_del_mcast(struct ti_softc *, struct ether_addr *);
static void ti_setmulti(struct ti_softc *);
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
static void ti_mem_read(struct ti_softc *, uint32_t, uint32_t, void *);
static void ti_mem_write(struct ti_softc *, uint32_t, uint32_t, void *);
static void ti_mem_zero(struct ti_softc *, uint32_t, uint32_t);
static int ti_copy_mem(struct ti_softc *, uint32_t, uint32_t, caddr_t, int,
style. No functional changes.
2011-11-04 18:28:10 +00:00
int);
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
static int ti_copy_scratch(struct ti_softc *, uint32_t, uint32_t, caddr_t,
style. No functional changes.
2011-11-04 18:28:10 +00:00
int, int, int);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int ti_bcopy_swap(const void *, void *, size_t, ti_swap_type);
static void ti_loadfw(struct ti_softc *);
static void ti_cmd(struct ti_softc *, struct ti_cmd_desc *);
static void ti_cmd_ext(struct ti_softc *, struct ti_cmd_desc *, caddr_t, int);
static void ti_handle_events(struct ti_softc *);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
static void ti_dma_map_addr(void *, bus_dma_segment_t *, int, int);
static int ti_dma_alloc(struct ti_softc *);
static void ti_dma_free(struct ti_softc *);
static int ti_dma_ring_alloc(struct ti_softc *, bus_size_t, bus_size_t,
bus_dma_tag_t *, uint8_t **, bus_dmamap_t *, bus_addr_t *, const char *);
static void ti_dma_ring_free(struct ti_softc *, bus_dma_tag_t *, uint8_t **,
bus_dmamap_t *);
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
static int ti_newbuf_std(struct ti_softc *, int);
static int ti_newbuf_mini(struct ti_softc *, int);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int ti_newbuf_jumbo(struct ti_softc *, int, struct mbuf *);
static int ti_init_rx_ring_std(struct ti_softc *);
static void ti_free_rx_ring_std(struct ti_softc *);
static int ti_init_rx_ring_jumbo(struct ti_softc *);
static void ti_free_rx_ring_jumbo(struct ti_softc *);
static int ti_init_rx_ring_mini(struct ti_softc *);
static void ti_free_rx_ring_mini(struct ti_softc *);
static void ti_free_tx_ring(struct ti_softc *);
static int ti_init_tx_ring(struct ti_softc *);
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
static void ti_discard_std(struct ti_softc *, int);
#ifndef TI_SF_BUF_JUMBO
static void ti_discard_jumbo(struct ti_softc *, int);
#endif
static void ti_discard_mini(struct ti_softc *, int);
Fix style(9) issues with __P removal. Noticed by: bde
2005-02-24 22:33:05 +00:00
static int ti_64bitslot_war(struct ti_softc *);
static int ti_chipinit(struct ti_softc *);
static int ti_gibinit(struct ti_softc *);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#ifdef TI_JUMBO_HDRSPLIT
style. No functional changes.
2011-11-04 18:28:10 +00:00
static __inline void ti_hdr_split(struct mbuf *top, int hdr_len, int pkt_len,
int idx);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#endif /* TI_JUMBO_HDRSPLIT */
Export sysctl node for various interrupt moderation parameters and have administrators control them. ti(4) provides a character device to control various other features of driver via ioctls but users had to write their own code to manipulate these parameters. It seems some default values for these parameters are not optimal on today's system but leave it as it was and let administrators change them. The following parameters could be changed: dev.ti.%d.rx_coal_ticks dev.ti.%d.rx_max_coal_bds dev.ti.%d.tx_coal_ticks dev.ti.%d.tx_max_coal_bds dev.ti.%d.tx_buf_ratio dev.ti.%d.stat_ticks The interface has to be brought down and up again before a change takes effect. ti(4) controller supports hardware MAC counters with additional DMA statistics. So it's doable to export these counters via sysctl interface. Unfortunately, these counters are cumulative such that driver have to either send an explicit clear command to controller after extracting them or have to maintain internal counters to get actual changes. Neither look good to me so counters were not exported via sysctl.
2011-11-14 19:10:20 +00:00
static void ti_sysctl_node(struct ti_softc *);
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
static device_method_t ti_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, ti_probe),
DEVMETHOD(device_attach, ti_attach),
DEVMETHOD(device_detach, ti_detach),
DEVMETHOD(device_shutdown, ti_shutdown),
{ 0, 0 }
};
static driver_t ti_driver = {
Grrr. Okay, changing the devnames was a bad idea. Put them back the way they were.
1999-09-20 08:47:11 +00:00
"ti",
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
ti_methods,
sizeof(struct ti_softc)
};
static devclass_t ti_devclass;
- Express hard dependencies on bus (pci, isa, pccard) and network layer (ether). - Don't abuse module names to facilitate ifconfig module loading; such abuse isn't really needed. (And if we do need type information associated with a module then we should make it explicit and not use hacks.)
2003-04-15 06:37:30 +00:00
DRIVER_MODULE(ti, pci, ti_driver, ti_devclass, 0, 0);
MODULE_DEPEND(ti, pci, 1, 1, 1);
MODULE_DEPEND(ti, ether, 1, 1, 1);
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Send an instruction or address to the EEPROM, check for ACK.
*/
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
static uint32_t
style. No functional changes.
2011-11-04 18:28:10 +00:00
ti_eeprom_putbyte(struct ti_softc *sc, int byte)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int i, ack = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Make sure we're in TX mode.
*/
TI_SETBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_TXEN);
/*
* Feed in each bit and stobe the clock.
*/
for (i = 0x80; i; i >>= 1) {
if (byte & i) {
TI_SETBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_DOUT);
} else {
TI_CLRBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_DOUT);
}
DELAY(1);
TI_SETBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_CLK);
DELAY(1);
TI_CLRBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_CLK);
}
/*
* Turn off TX mode.
*/
TI_CLRBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_TXEN);
/*
* Check for ack.
*/
TI_SETBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_CLK);
ack = CSR_READ_4(sc, TI_MISC_LOCAL_CTL) & TI_MLC_EE_DIN;
TI_CLRBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_CLK);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ack);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Read a byte of data stored in the EEPROM at address 'addr.'
* We have to send two address bytes since the EEPROM can hold
* more than 256 bytes of data.
*/
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
static uint8_t
ti_eeprom_getbyte(struct ti_softc *sc, int addr, uint8_t *dest)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int i;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint8_t byte = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
EEPROM_START;
/*
* Send write control code to EEPROM.
*/
if (ti_eeprom_putbyte(sc, EEPROM_CTL_WRITE)) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev,
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"failed to send write command, status: %x\n",
CSR_READ_4(sc, TI_MISC_LOCAL_CTL));
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (1);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Send first byte of address of byte we want to read.
*/
if (ti_eeprom_putbyte(sc, (addr >> 8) & 0xFF)) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "failed to send address, status: %x\n",
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
CSR_READ_4(sc, TI_MISC_LOCAL_CTL));
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (1);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Send second byte address of byte we want to read.
*/
if (ti_eeprom_putbyte(sc, addr & 0xFF)) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "failed to send address, status: %x\n",
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
CSR_READ_4(sc, TI_MISC_LOCAL_CTL));
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (1);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
EEPROM_STOP;
EEPROM_START;
/*
* Send read control code to EEPROM.
*/
if (ti_eeprom_putbyte(sc, EEPROM_CTL_READ)) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev,
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"failed to send read command, status: %x\n",
CSR_READ_4(sc, TI_MISC_LOCAL_CTL));
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (1);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Start reading bits from EEPROM.
*/
TI_CLRBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_TXEN);
for (i = 0x80; i; i >>= 1) {
TI_SETBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_CLK);
DELAY(1);
if (CSR_READ_4(sc, TI_MISC_LOCAL_CTL) & TI_MLC_EE_DIN)
byte |= i;
TI_CLRBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_EE_CLK);
DELAY(1);
}
EEPROM_STOP;
/*
* No ACK generated for read, so just return byte.
*/
*dest = byte;
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Read a sequence of bytes from the EEPROM.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_read_eeprom(struct ti_softc *sc, caddr_t dest, int off, int cnt)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int err = 0, i;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint8_t byte = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
for (i = 0; i < cnt; i++) {
err = ti_eeprom_getbyte(sc, off + i, &byte);
if (err)
break;
*(dest + i) = byte;
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (err ? 1 : 0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
* NIC memory read function.
* Can be used to copy data from NIC local memory.
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
ti_mem_read(struct ti_softc *sc, uint32_t addr, uint32_t len, void *buf)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int segptr, segsize, cnt;
char *ptr;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
segptr = addr;
cnt = len;
ptr = buf;
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
while (cnt) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (cnt < TI_WINLEN)
segsize = cnt;
else
segsize = TI_WINLEN - (segptr % TI_WINLEN);
CSR_WRITE_4(sc, TI_WINBASE, (segptr & ~(TI_WINLEN - 1)));
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
bus_space_read_region_4(sc->ti_btag, sc->ti_bhandle,
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
TI_WINDOW + (segptr & (TI_WINLEN - 1)), (uint32_t *)ptr,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
segsize / 4);
ptr += segsize;
segptr += segsize;
cnt -= segsize;
}
}
/*
* NIC memory write function.
* Can be used to copy data into NIC local memory.
*/
static void
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
ti_mem_write(struct ti_softc *sc, uint32_t addr, uint32_t len, void *buf)
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int segptr, segsize, cnt;
char *ptr;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
segptr = addr;
cnt = len;
ptr = buf;
while (cnt) {
if (cnt < TI_WINLEN)
segsize = cnt;
else
segsize = TI_WINLEN - (segptr % TI_WINLEN);
CSR_WRITE_4(sc, TI_WINBASE, (segptr & ~(TI_WINLEN - 1)));
bus_space_write_region_4(sc->ti_btag, sc->ti_bhandle,
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
TI_WINDOW + (segptr & (TI_WINLEN - 1)), (uint32_t *)ptr,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
segsize / 4);
ptr += segsize;
segptr += segsize;
cnt -= segsize;
}
}
/*
* NIC memory read function.
* Can be used to clear a section of NIC local memory.
*/
static void
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
ti_mem_zero(struct ti_softc *sc, uint32_t addr, uint32_t len)
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int segptr, segsize, cnt;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
segptr = addr;
cnt = len;
while (cnt) {
if (cnt < TI_WINLEN)
segsize = cnt;
else
segsize = TI_WINLEN - (segptr % TI_WINLEN);
CSR_WRITE_4(sc, TI_WINBASE, (segptr & ~(TI_WINLEN - 1)));
bus_space_set_region_4(sc->ti_btag, sc->ti_bhandle,
TI_WINDOW + (segptr & (TI_WINLEN - 1)), 0, segsize / 4);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
segptr += segsize;
cnt -= segsize;
}
}
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
static int
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
ti_copy_mem(struct ti_softc *sc, uint32_t tigon_addr, uint32_t len,
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
caddr_t buf, int useraddr, int readdata)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int segptr, segsize, cnt;
caddr_t ptr;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t origwin;
style. No functional changes.
2011-11-04 18:28:10 +00:00
int resid, segresid;
int first_pass;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK_ASSERT(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* At the moment, we don't handle non-aligned cases, we just bail.
* If this proves to be a problem, it will be fixed.
*/
style. No functional changes.
2011-11-10 22:15:11 +00:00
if (readdata == 0 && (tigon_addr & 0x3) != 0) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "%s: tigon address %#x isn't "
"word-aligned\n", __func__, tigon_addr);
device_printf(sc->ti_dev, "%s: unaligned writes aren't "
"yet supported\n", __func__);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (EINVAL);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
segptr = tigon_addr & ~0x3;
segresid = tigon_addr - segptr;
/*
* This is the non-aligned amount left over that we'll need to
* copy.
*/
resid = len & 0x3;
/* Add in the left over amount at the front of the buffer */
resid += segresid;
cnt = len & ~0x3;
/*
* If resid + segresid is >= 4, add multiples of 4 to the count and
* decrease the residual by that much.
*/
cnt += resid & ~0x3;
resid -= resid & ~0x3;
ptr = buf;
first_pass = 1;
/*
* Save the old window base value.
*/
origwin = CSR_READ_4(sc, TI_WINBASE);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
while (cnt) {
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
bus_size_t ti_offset;
if (cnt < TI_WINLEN)
segsize = cnt;
else
segsize = TI_WINLEN - (segptr % TI_WINLEN);
CSR_WRITE_4(sc, TI_WINBASE, (segptr & ~(TI_WINLEN - 1)));
ti_offset = TI_WINDOW + (segptr & (TI_WINLEN -1));
if (readdata) {
style. No functional changes.
2011-11-10 22:15:11 +00:00
bus_space_read_region_4(sc->ti_btag, sc->ti_bhandle,
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
ti_offset, (uint32_t *)sc->ti_membuf, segsize >> 2);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (useraddr) {
/*
* Yeah, this is a little on the kludgy
* side, but at least this code is only
* used for debugging.
*/
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
ti_bcopy_swap(sc->ti_membuf, sc->ti_membuf2,
segsize, TI_SWAP_NTOH);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (first_pass) {
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
copyout(&sc->ti_membuf2[segresid], ptr,
style. No functional changes.
2011-11-10 22:15:11 +00:00
segsize - segresid);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
first_pass = 0;
} else
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
copyout(sc->ti_membuf2, ptr, segsize);
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
} else {
if (first_pass) {
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
ti_bcopy_swap(sc->ti_membuf,
sc->ti_membuf2, segsize,
TI_SWAP_NTOH);
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
bcopy(&sc->ti_membuf2[segresid], ptr,
style. No functional changes.
2011-11-10 22:15:11 +00:00
segsize - segresid);
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
first_pass = 0;
} else
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
ti_bcopy_swap(sc->ti_membuf, ptr,
segsize, TI_SWAP_NTOH);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
} else {
if (useraddr) {
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
copyin(ptr, sc->ti_membuf2, segsize);
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
ti_bcopy_swap(sc->ti_membuf2, sc->ti_membuf,
segsize, TI_SWAP_HTON);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
} else
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
ti_bcopy_swap(ptr, sc->ti_membuf, segsize,
style. No functional changes.
2011-11-10 22:15:11 +00:00
TI_SWAP_HTON);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
style. No functional changes.
2011-11-10 22:15:11 +00:00
bus_space_write_region_4(sc->ti_btag, sc->ti_bhandle,
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
ti_offset, (uint32_t *)sc->ti_membuf, segsize >> 2);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
segptr += segsize;
ptr += segsize;
cnt -= segsize;
}
/*
* Handle leftover, non-word-aligned bytes.
*/
if (resid != 0) {
style. No functional changes.
2011-11-10 22:15:11 +00:00
uint32_t tmpval, tmpval2;
bus_size_t ti_offset;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* Set the segment pointer.
*/
CSR_WRITE_4(sc, TI_WINBASE, (segptr & ~(TI_WINLEN - 1)));
ti_offset = TI_WINDOW + (segptr & (TI_WINLEN - 1));
/*
* First, grab whatever is in our source/destination.
* We'll obviously need this for reads, but also for
* writes, since we'll be doing read/modify/write.
*/
bus_space_read_region_4(sc->ti_btag, sc->ti_bhandle,
style. No functional changes.
2011-11-10 22:15:11 +00:00
ti_offset, &tmpval, 1);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* Next, translate this from little-endian to big-endian
* (at least on i386 boxes).
*/
tmpval2 = ntohl(tmpval);
if (readdata) {
/*
* If we're reading, just copy the leftover number
* of bytes from the host byte order buffer to
* the user's buffer.
*/
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
if (useraddr) {
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
copyout(&tmpval2, ptr, resid);
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
} else
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
bcopy(&tmpval2, ptr, resid);
} else {
/*
* If we're writing, first copy the bytes to be
* written into the network byte order buffer,
* leaving the rest of the buffer with whatever was
* originally in there. Then, swap the bytes
* around into host order and write them out.
*
* XXX KDM the read side of this has been verified
* to work, but the write side of it has not been
* verified. So user beware.
*/
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
if (useraddr) {
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
copyin(ptr, &tmpval2, resid);
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
} else
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
bcopy(ptr, &tmpval2, resid);
tmpval = htonl(tmpval2);
bus_space_write_region_4(sc->ti_btag, sc->ti_bhandle,
style. No functional changes.
2011-11-10 22:15:11 +00:00
ti_offset, &tmpval, 1);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
}
CSR_WRITE_4(sc, TI_WINBASE, origwin);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
static int
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
ti_copy_scratch(struct ti_softc *sc, uint32_t tigon_addr, uint32_t len,
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
caddr_t buf, int useraddr, int readdata, int cpu)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
{
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t segptr;
style. No functional changes.
2011-11-04 18:28:10 +00:00
int cnt;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t tmpval, tmpval2;
style. No functional changes.
2011-11-04 18:28:10 +00:00
caddr_t ptr;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK_ASSERT(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* At the moment, we don't handle non-aligned cases, we just bail.
* If this proves to be a problem, it will be fixed.
*/
if (tigon_addr & 0x3) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "%s: tigon address %#x "
"isn't word-aligned\n", __func__, tigon_addr);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (EINVAL);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
if (len & 0x3) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "%s: transfer length %d "
"isn't word-aligned\n", __func__, len);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (EINVAL);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
segptr = tigon_addr;
cnt = len;
ptr = buf;
while (cnt) {
CSR_WRITE_4(sc, CPU_REG(TI_SRAM_ADDR, cpu), segptr);
if (readdata) {
tmpval2 = CSR_READ_4(sc, CPU_REG(TI_SRAM_DATA, cpu));
tmpval = ntohl(tmpval2);
/*
* Note: I've used this debugging interface
* extensively with Alteon's 12.3.15 firmware,
* compiled with GCC 2.7.2.1 and binutils 2.9.1.
*
* When you compile the firmware without
* optimization, which is necessary sometimes in
* order to properly step through it, you sometimes
Whitespace nits
2004-07-05 22:35:18 +00:00
* read out a bogus value of 0xc0017c instead of
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
* whatever was supposed to be in that scratchpad
* location. That value is on the stack somewhere,
* but I've never been able to figure out what was
* causing the problem.
*
* The address seems to pop up in random places,
* often not in the same place on two subsequent
* reads.
*
* In any case, the underlying data doesn't seem
* to be affected, just the value read out.
*
* KDM, 3/7/2000
*/
if (tmpval2 == 0xc0017c)
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "found 0xc0017c at "
"%#x (tmpval2)\n", segptr);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (tmpval == 0xc0017c)
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "found 0xc0017c at "
"%#x (tmpval)\n", segptr);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (useraddr)
copyout(&tmpval, ptr, 4);
else
bcopy(&tmpval, ptr, 4);
} else {
if (useraddr)
copyin(ptr, &tmpval2, 4);
else
bcopy(ptr, &tmpval2, 4);
tmpval = htonl(tmpval2);
CSR_WRITE_4(sc, CPU_REG(TI_SRAM_DATA, cpu), tmpval);
}
cnt -= 4;
segptr += 4;
ptr += 4;
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_bcopy_swap(const void *src, void *dst, size_t len, ti_swap_type swap_type)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
{
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
const uint8_t *tmpsrc;
uint8_t *tmpdst;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
size_t tmplen;
if (len & 0x3) {
style. No functional changes.
2011-11-10 22:15:11 +00:00
printf("ti_bcopy_swap: length %zd isn't 32-bit aligned\n", len);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (-1);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
tmpsrc = src;
tmpdst = dst;
tmplen = len;
while (tmplen) {
if (swap_type == TI_SWAP_NTOH)
style. No functional changes.
2011-11-10 22:15:11 +00:00
*(uint32_t *)tmpdst = ntohl(*(const uint32_t *)tmpsrc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
else
style. No functional changes.
2011-11-10 22:15:11 +00:00
*(uint32_t *)tmpdst = htonl(*(const uint32_t *)tmpsrc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
tmpsrc += 4;
tmpdst += 4;
tmplen -= 4;
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Load firmware image into the NIC. Check that the firmware revision
* is acceptable and see if we want the firmware for the Tigon 1 or
* Tigon 2.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_loadfw(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK_ASSERT(sc);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
switch (sc->ti_hwrev) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
case TI_HWREV_TIGON:
if (tigonFwReleaseMajor != TI_FIRMWARE_MAJOR ||
tigonFwReleaseMinor != TI_FIRMWARE_MINOR ||
tigonFwReleaseFix != TI_FIRMWARE_FIX) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "firmware revision mismatch; "
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"want %d.%d.%d, got %d.%d.%d\n",
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_FIRMWARE_MAJOR, TI_FIRMWARE_MINOR,
TI_FIRMWARE_FIX, tigonFwReleaseMajor,
tigonFwReleaseMinor, tigonFwReleaseFix);
return;
}
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
ti_mem_write(sc, tigonFwTextAddr, tigonFwTextLen, tigonFwText);
ti_mem_write(sc, tigonFwDataAddr, tigonFwDataLen, tigonFwData);
ti_mem_write(sc, tigonFwRodataAddr, tigonFwRodataLen,
tigonFwRodata);
ti_mem_zero(sc, tigonFwBssAddr, tigonFwBssLen);
ti_mem_zero(sc, tigonFwSbssAddr, tigonFwSbssLen);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_CPU_PROGRAM_COUNTER, tigonFwStartAddr);
break;
case TI_HWREV_TIGON_II:
if (tigon2FwReleaseMajor != TI_FIRMWARE_MAJOR ||
tigon2FwReleaseMinor != TI_FIRMWARE_MINOR ||
tigon2FwReleaseFix != TI_FIRMWARE_FIX) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "firmware revision mismatch; "
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"want %d.%d.%d, got %d.%d.%d\n",
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_FIRMWARE_MAJOR, TI_FIRMWARE_MINOR,
TI_FIRMWARE_FIX, tigon2FwReleaseMajor,
tigon2FwReleaseMinor, tigon2FwReleaseFix);
return;
}
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
ti_mem_write(sc, tigon2FwTextAddr, tigon2FwTextLen,
tigon2FwText);
ti_mem_write(sc, tigon2FwDataAddr, tigon2FwDataLen,
tigon2FwData);
ti_mem_write(sc, tigon2FwRodataAddr, tigon2FwRodataLen,
tigon2FwRodata);
ti_mem_zero(sc, tigon2FwBssAddr, tigon2FwBssLen);
ti_mem_zero(sc, tigon2FwSbssAddr, tigon2FwSbssLen);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_CPU_PROGRAM_COUNTER, tigon2FwStartAddr);
break;
default:
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev,
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"can't load firmware: unknown hardware rev\n");
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
}
}
/*
* Send the NIC a command via the command ring.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_cmd(struct ti_softc *sc, struct ti_cmd_desc *cmd)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int index;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
index = sc->ti_cmd_saved_prodidx;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
CSR_WRITE_4(sc, TI_GCR_CMDRING + (index * 4), *(uint32_t *)(cmd));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_INC(index, TI_CMD_RING_CNT);
CSR_WRITE_4(sc, TI_MB_CMDPROD_IDX, index);
sc->ti_cmd_saved_prodidx = index;
}
/*
* Send the NIC an extended command. The 'len' parameter specifies the
* number of command slots to include after the initial command.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_cmd_ext(struct ti_softc *sc, struct ti_cmd_desc *cmd, caddr_t arg, int len)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int index;
int i;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
index = sc->ti_cmd_saved_prodidx;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
CSR_WRITE_4(sc, TI_GCR_CMDRING + (index * 4), *(uint32_t *)(cmd));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_INC(index, TI_CMD_RING_CNT);
for (i = 0; i < len; i++) {
CSR_WRITE_4(sc, TI_GCR_CMDRING + (index * 4),
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
*(uint32_t *)(&arg[i * 4]));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_INC(index, TI_CMD_RING_CNT);
}
CSR_WRITE_4(sc, TI_MB_CMDPROD_IDX, index);
sc->ti_cmd_saved_prodidx = index;
}
/*
* Handle events that have triggered interrupts.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_handle_events(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_event_desc *e;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_rdata.ti_event_ring == NULL)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
return;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_event_ring_tag,
sc->ti_cdata.ti_event_ring_map, BUS_DMASYNC_POSTREAD);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
while (sc->ti_ev_saved_considx != sc->ti_ev_prodidx.ti_idx) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
e = &sc->ti_rdata.ti_event_ring[sc->ti_ev_saved_considx];
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
switch (TI_EVENT_EVENT(e)) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
case TI_EV_LINKSTAT_CHANGED:
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
sc->ti_linkstat = TI_EVENT_CODE(e);
Announce IFCAP_LINKSTATE capability and let network stack know link state changes. Hide superfluous link up/down message under bootverbose since if_link_state_change(9) shows that information. While I'm here, change baudrate with the resolved speed of the established link instead of blindly setting it 1G. Unfortunately, it seems there is no way to differentiate 10/100Mbps from non-gigabit link so just assume we established a 100Mbps link if current link is not a gigabit link.
2011-11-04 20:43:37 +00:00
if (sc->ti_linkstat == TI_EV_CODE_LINK_UP) {
if_link_state_change(sc->ti_ifp, LINK_STATE_UP);
sc->ti_ifp->if_baudrate = IF_Mbps(100);
if (bootverbose)
device_printf(sc->ti_dev,
"10/100 link up\n");
} else if (sc->ti_linkstat == TI_EV_CODE_GIG_LINK_UP) {
if_link_state_change(sc->ti_ifp, LINK_STATE_UP);
sc->ti_ifp->if_baudrate = IF_Gbps(1UL);
if (bootverbose)
device_printf(sc->ti_dev,
"gigabit link up\n");
} else if (sc->ti_linkstat == TI_EV_CODE_LINK_DOWN) {
if_link_state_change(sc->ti_ifp,
LINK_STATE_DOWN);
sc->ti_ifp->if_baudrate = 0;
if (bootverbose)
device_printf(sc->ti_dev,
"link down\n");
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
case TI_EV_ERROR:
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
if (TI_EVENT_CODE(e) == TI_EV_CODE_ERR_INVAL_CMD)
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "invalid command\n");
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
else if (TI_EVENT_CODE(e) == TI_EV_CODE_ERR_UNIMP_CMD)
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "unknown command\n");
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
else if (TI_EVENT_CODE(e) == TI_EV_CODE_ERR_BADCFG)
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "bad config data\n");
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
case TI_EV_FIRMWARE_UP:
ti_init2(sc);
break;
case TI_EV_STATS_UPDATED:
ti_stats_update(sc);
break;
case TI_EV_RESET_JUMBO_RING:
case TI_EV_MCAST_UPDATED:
/* Who cares. */
break;
default:
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "unknown event: %d\n",
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_EVENT_EVENT(e));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
}
/* Advance the consumer index. */
TI_INC(sc->ti_ev_saved_considx, TI_EVENT_RING_CNT);
CSR_WRITE_4(sc, TI_GCR_EVENTCONS_IDX, sc->ti_ev_saved_considx);
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_event_ring_tag,
sc->ti_cdata.ti_event_ring_map, BUS_DMASYNC_PREREAD);
}
struct ti_dmamap_arg {
bus_addr_t ti_busaddr;
};
static void
ti_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
struct ti_dmamap_arg *ctx;
if (error)
return;
KASSERT(nseg == 1, ("%s: %d segments returned!", __func__, nseg));
ctx = arg;
ctx->ti_busaddr = segs->ds_addr;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
static int
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
ti_dma_ring_alloc(struct ti_softc *sc, bus_size_t alignment, bus_size_t maxsize,
bus_dma_tag_t *tag, uint8_t **ring, bus_dmamap_t *map, bus_addr_t *paddr,
const char *msg)
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
{
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
struct ti_dmamap_arg ctx;
int error;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dma_tag_create(sc->ti_cdata.ti_parent_tag,
alignment, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, maxsize, 1, maxsize, 0, NULL, NULL, tag);
if (error != 0) {
device_printf(sc->ti_dev,
"could not create %s dma tag\n", msg);
return (error);
}
/* Allocate DMA'able memory for ring. */
error = bus_dmamem_alloc(*tag, (void **)ring,
BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT, map);
if (error != 0) {
device_printf(sc->ti_dev,
"could not allocate DMA'able memory for %s\n", msg);
return (error);
}
/* Load the address of the ring. */
ctx.ti_busaddr = 0;
error = bus_dmamap_load(*tag, *map, *ring, maxsize, ti_dma_map_addr,
&ctx, BUS_DMA_NOWAIT);
if (error != 0) {
device_printf(sc->ti_dev,
"could not load DMA'able memory for %s\n", msg);
return (error);
}
*paddr = ctx.ti_busaddr;
return (0);
}
static void
ti_dma_ring_free(struct ti_softc *sc, bus_dma_tag_t *tag, uint8_t **ring,
bus_dmamap_t *map)
{
if (*map != NULL)
bus_dmamap_unload(*tag, *map);
if (*map != NULL && *ring != NULL) {
bus_dmamem_free(*tag, *ring, *map);
*ring = NULL;
*map = NULL;
}
if (*tag) {
bus_dma_tag_destroy(*tag);
*tag = NULL;
}
}
static int
ti_dma_alloc(struct ti_softc *sc)
{
bus_addr_t lowaddr;
int i, error;
lowaddr = BUS_SPACE_MAXADDR;
if (sc->ti_dac == 0)
lowaddr = BUS_SPACE_MAXADDR_32BIT;
error = bus_dma_tag_create(bus_get_dma_tag(sc->ti_dev), 1, 0, lowaddr,
BUS_SPACE_MAXADDR, NULL, NULL, BUS_SPACE_MAXSIZE_32BIT, 0,
BUS_SPACE_MAXSIZE_32BIT, 0, NULL, NULL,
&sc->ti_cdata.ti_parent_tag);
if (error != 0) {
device_printf(sc->ti_dev,
"could not allocate parent dma tag\n");
return (ENOMEM);
}
error = ti_dma_ring_alloc(sc, TI_RING_ALIGN, sizeof(struct ti_gib),
&sc->ti_cdata.ti_gib_tag, (uint8_t **)&sc->ti_rdata.ti_info,
&sc->ti_cdata.ti_gib_map, &sc->ti_rdata.ti_info_paddr, "GIB");
if (error)
return (error);
/* Producer/consumer status */
error = ti_dma_ring_alloc(sc, TI_RING_ALIGN, sizeof(struct ti_status),
&sc->ti_cdata.ti_status_tag, (uint8_t **)&sc->ti_rdata.ti_status,
&sc->ti_cdata.ti_status_map, &sc->ti_rdata.ti_status_paddr,
"event ring");
if (error)
return (error);
/* Event ring */
error = ti_dma_ring_alloc(sc, TI_RING_ALIGN, TI_EVENT_RING_SZ,
&sc->ti_cdata.ti_event_ring_tag,
(uint8_t **)&sc->ti_rdata.ti_event_ring,
&sc->ti_cdata.ti_event_ring_map, &sc->ti_rdata.ti_event_ring_paddr,
"event ring");
if (error)
return (error);
/* Command ring lives in shared memory so no need to create DMA area. */
/* Standard RX ring */
error = ti_dma_ring_alloc(sc, TI_RING_ALIGN, TI_STD_RX_RING_SZ,
&sc->ti_cdata.ti_rx_std_ring_tag,
(uint8_t **)&sc->ti_rdata.ti_rx_std_ring,
&sc->ti_cdata.ti_rx_std_ring_map,
&sc->ti_rdata.ti_rx_std_ring_paddr, "RX ring");
if (error)
return (error);
/* Jumbo RX ring */
error = ti_dma_ring_alloc(sc, TI_JUMBO_RING_ALIGN, TI_JUMBO_RX_RING_SZ,
&sc->ti_cdata.ti_rx_jumbo_ring_tag,
(uint8_t **)&sc->ti_rdata.ti_rx_jumbo_ring,
&sc->ti_cdata.ti_rx_jumbo_ring_map,
&sc->ti_rdata.ti_rx_jumbo_ring_paddr, "jumbo RX ring");
if (error)
return (error);
/* RX return ring */
error = ti_dma_ring_alloc(sc, TI_RING_ALIGN, TI_RX_RETURN_RING_SZ,
&sc->ti_cdata.ti_rx_return_ring_tag,
(uint8_t **)&sc->ti_rdata.ti_rx_return_ring,
&sc->ti_cdata.ti_rx_return_ring_map,
&sc->ti_rdata.ti_rx_return_ring_paddr, "RX return ring");
if (error)
return (error);
/* Create DMA tag for standard RX mbufs. */
error = bus_dma_tag_create(sc->ti_cdata.ti_parent_tag, 1, 0,
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, MCLBYTES, 1,
MCLBYTES, 0, NULL, NULL, &sc->ti_cdata.ti_rx_std_tag);
if (error) {
device_printf(sc->ti_dev, "could not allocate RX dma tag\n");
return (error);
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Create DMA tag for jumbo RX mbufs. */
#ifdef TI_SF_BUF_JUMBO
/*
* The VM system will take care of providing aligned pages. Alignment
* is set to 1 here so that busdma resources won't be wasted.
*/
error = bus_dma_tag_create(sc->ti_cdata.ti_parent_tag, 1, 0,
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, PAGE_SIZE * 4, 4,
PAGE_SIZE, 0, NULL, NULL, &sc->ti_cdata.ti_rx_jumbo_tag);
#else
error = bus_dma_tag_create(sc->ti_cdata.ti_parent_tag, 1, 0,
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, MJUM9BYTES, 1,
MJUM9BYTES, 0, NULL, NULL, &sc->ti_cdata.ti_rx_jumbo_tag);
#endif
if (error) {
device_printf(sc->ti_dev,
"could not allocate jumbo RX dma tag\n");
return (error);
}
/* Create DMA tag for TX mbufs. */
error = bus_dma_tag_create(sc->ti_cdata.ti_parent_tag, 1,
0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
MCLBYTES * TI_MAXTXSEGS, TI_MAXTXSEGS, MCLBYTES, 0, NULL, NULL,
&sc->ti_cdata.ti_tx_tag);
if (error) {
device_printf(sc->ti_dev, "could not allocate TX dma tag\n");
return (ENOMEM);
}
/* Create DMA maps for RX buffers. */
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
for (i = 0; i < TI_STD_RX_RING_CNT; i++) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_create(sc->ti_cdata.ti_rx_std_tag, 0,
&sc->ti_cdata.ti_rx_std_maps[i]);
if (error) {
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
device_printf(sc->ti_dev,
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
"could not create DMA map for RX\n");
return (error);
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_create(sc->ti_cdata.ti_rx_std_tag, 0,
&sc->ti_cdata.ti_rx_std_sparemap);
if (error) {
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
device_printf(sc->ti_dev,
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
"could not create spare DMA map for RX\n");
return (error);
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Create DMA maps for jumbo RX buffers. */
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
for (i = 0; i < TI_JUMBO_RX_RING_CNT; i++) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_create(sc->ti_cdata.ti_rx_jumbo_tag, 0,
&sc->ti_cdata.ti_rx_jumbo_maps[i]);
if (error) {
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
device_printf(sc->ti_dev,
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
"could not create DMA map for jumbo RX\n");
return (error);
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_create(sc->ti_cdata.ti_rx_jumbo_tag, 0,
&sc->ti_cdata.ti_rx_jumbo_sparemap);
if (error) {
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
device_printf(sc->ti_dev,
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
"could not create spare DMA map for jumbo RX\n");
return (error);
}
/* Create DMA maps for TX buffers. */
for (i = 0; i < TI_TX_RING_CNT; i++) {
error = bus_dmamap_create(sc->ti_cdata.ti_tx_tag, 0,
&sc->ti_cdata.ti_txdesc[i].tx_dmamap);
if (error) {
device_printf(sc->ti_dev,
"could not create DMA map for TX\n");
return (ENOMEM);
}
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
}
Mini ring is not available on Tigon 1 so do not create DMA maps for mini ring on Tigon 1 to save resources.
2011-11-07 22:17:44 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Mini ring and TX ring is not available on Tigon 1. */
Mini ring is not available on Tigon 1 so do not create DMA maps for mini ring on Tigon 1 to save resources.
2011-11-07 22:17:44 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON)
return (0);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* TX ring */
error = ti_dma_ring_alloc(sc, TI_RING_ALIGN, TI_TX_RING_SZ,
&sc->ti_cdata.ti_tx_ring_tag, (uint8_t **)&sc->ti_rdata.ti_tx_ring,
&sc->ti_cdata.ti_tx_ring_map, &sc->ti_rdata.ti_tx_ring_paddr,
"TX ring");
if (error)
return (error);
/* Mini RX ring */
error = ti_dma_ring_alloc(sc, TI_RING_ALIGN, TI_MINI_RX_RING_SZ,
&sc->ti_cdata.ti_rx_mini_ring_tag,
(uint8_t **)&sc->ti_rdata.ti_rx_mini_ring,
&sc->ti_cdata.ti_rx_mini_ring_map,
&sc->ti_rdata.ti_rx_mini_ring_paddr, "mini RX ring");
if (error)
return (error);
/* Create DMA tag for mini RX mbufs. */
error = bus_dma_tag_create(sc->ti_cdata.ti_parent_tag, 1, 0,
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, MHLEN, 1,
MHLEN, 0, NULL, NULL, &sc->ti_cdata.ti_rx_mini_tag);
if (error) {
device_printf(sc->ti_dev,
"could not allocate mini RX dma tag\n");
return (error);
}
/* Create DMA maps for mini RX buffers. */
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
for (i = 0; i < TI_MINI_RX_RING_CNT; i++) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_create(sc->ti_cdata.ti_rx_mini_tag, 0,
&sc->ti_cdata.ti_rx_mini_maps[i]);
if (error) {
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
device_printf(sc->ti_dev,
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
"could not create DMA map for mini RX\n");
return (error);
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_create(sc->ti_cdata.ti_rx_mini_tag, 0,
&sc->ti_cdata.ti_rx_mini_sparemap);
if (error) {
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
device_printf(sc->ti_dev,
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
"could not create spare DMA map for mini RX\n");
return (error);
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
}
return (0);
}
static void
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
ti_dma_free(struct ti_softc *sc)
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
{
int i;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Destroy DMA maps for RX buffers. */
for (i = 0; i < TI_STD_RX_RING_CNT; i++) {
if (sc->ti_cdata.ti_rx_std_maps[i]) {
bus_dmamap_destroy(sc->ti_cdata.ti_rx_std_tag,
sc->ti_cdata.ti_rx_std_maps[i]);
sc->ti_cdata.ti_rx_std_maps[i] = NULL;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_cdata.ti_rx_std_sparemap) {
bus_dmamap_destroy(sc->ti_cdata.ti_rx_std_tag,
sc->ti_cdata.ti_rx_std_sparemap);
sc->ti_cdata.ti_rx_std_sparemap = NULL;
}
if (sc->ti_cdata.ti_rx_std_tag) {
bus_dma_tag_destroy(sc->ti_cdata.ti_rx_std_tag);
sc->ti_cdata.ti_rx_std_tag = NULL;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
}
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Destroy DMA maps for jumbo RX buffers. */
for (i = 0; i < TI_JUMBO_RX_RING_CNT; i++) {
if (sc->ti_cdata.ti_rx_jumbo_maps[i]) {
bus_dmamap_destroy(sc->ti_cdata.ti_rx_jumbo_tag,
sc->ti_cdata.ti_rx_jumbo_maps[i]);
sc->ti_cdata.ti_rx_jumbo_maps[i] = NULL;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_cdata.ti_rx_jumbo_sparemap) {
bus_dmamap_destroy(sc->ti_cdata.ti_rx_jumbo_tag,
sc->ti_cdata.ti_rx_jumbo_sparemap);
sc->ti_cdata.ti_rx_jumbo_sparemap = NULL;
}
if (sc->ti_cdata.ti_rx_jumbo_tag) {
bus_dma_tag_destroy(sc->ti_cdata.ti_rx_jumbo_tag);
sc->ti_cdata.ti_rx_jumbo_tag = NULL;
}
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Destroy DMA maps for mini RX buffers. */
for (i = 0; i < TI_MINI_RX_RING_CNT; i++) {
if (sc->ti_cdata.ti_rx_mini_maps[i]) {
bus_dmamap_destroy(sc->ti_cdata.ti_rx_mini_tag,
sc->ti_cdata.ti_rx_mini_maps[i]);
sc->ti_cdata.ti_rx_mini_maps[i] = NULL;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_cdata.ti_rx_mini_sparemap) {
bus_dmamap_destroy(sc->ti_cdata.ti_rx_mini_tag,
sc->ti_cdata.ti_rx_mini_sparemap);
sc->ti_cdata.ti_rx_mini_sparemap = NULL;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_cdata.ti_rx_mini_tag) {
bus_dma_tag_destroy(sc->ti_cdata.ti_rx_mini_tag);
sc->ti_cdata.ti_rx_mini_tag = NULL;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Destroy DMA maps for TX buffers. */
for (i = 0; i < TI_TX_RING_CNT; i++) {
if (sc->ti_cdata.ti_txdesc[i].tx_dmamap) {
bus_dmamap_destroy(sc->ti_cdata.ti_tx_tag,
sc->ti_cdata.ti_txdesc[i].tx_dmamap);
sc->ti_cdata.ti_txdesc[i].tx_dmamap = NULL;
}
}
if (sc->ti_cdata.ti_tx_tag) {
bus_dma_tag_destroy(sc->ti_cdata.ti_tx_tag);
sc->ti_cdata.ti_tx_tag = NULL;
}
/* Destroy standard RX ring. */
ti_dma_ring_free(sc, &sc->ti_cdata.ti_rx_std_ring_tag,
(void *)&sc->ti_rdata.ti_rx_std_ring,
&sc->ti_cdata.ti_rx_std_ring_map);
/* Destroy jumbo RX ring. */
ti_dma_ring_free(sc, &sc->ti_cdata.ti_rx_jumbo_ring_tag,
(void *)&sc->ti_rdata.ti_rx_jumbo_ring,
&sc->ti_cdata.ti_rx_jumbo_ring_map);
/* Destroy mini RX ring. */
ti_dma_ring_free(sc, &sc->ti_cdata.ti_rx_mini_ring_tag,
(void *)&sc->ti_rdata.ti_rx_mini_ring,
&sc->ti_cdata.ti_rx_mini_ring_map);
/* Destroy RX return ring. */
ti_dma_ring_free(sc, &sc->ti_cdata.ti_rx_return_ring_tag,
(void *)&sc->ti_rdata.ti_rx_return_ring,
&sc->ti_cdata.ti_rx_return_ring_map);
/* Destroy TX ring. */
ti_dma_ring_free(sc, &sc->ti_cdata.ti_tx_ring_tag,
(void *)&sc->ti_rdata.ti_tx_ring, &sc->ti_cdata.ti_tx_ring_map);
/* Destroy status block. */
ti_dma_ring_free(sc, &sc->ti_cdata.ti_status_tag,
(void *)&sc->ti_rdata.ti_status, &sc->ti_cdata.ti_status_map);
/* Destroy event ring. */
ti_dma_ring_free(sc, &sc->ti_cdata.ti_event_ring_tag,
(void *)&sc->ti_rdata.ti_event_ring,
&sc->ti_cdata.ti_event_ring_map);
/* Destroy GIB */
ti_dma_ring_free(sc, &sc->ti_cdata.ti_gib_tag,
(void *)&sc->ti_rdata.ti_info, &sc->ti_cdata.ti_gib_map);
/* Destroy the parent tag. */
if (sc->ti_cdata.ti_parent_tag) {
bus_dma_tag_destroy(sc->ti_cdata.ti_parent_tag);
sc->ti_cdata.ti_parent_tag = NULL;
}
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Intialize a standard receive ring descriptor.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_newbuf_std(struct ti_softc *sc, int i)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
bus_dmamap_t map;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
bus_dma_segment_t segs[1];
struct mbuf *m;
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_rx_desc *r;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
int error, nsegs;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
if (m == NULL)
return (ENOBUFS);
m->m_len = m->m_pkthdr.len = MCLBYTES;
m_adj(m, ETHER_ALIGN);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_load_mbuf_sg(sc->ti_cdata.ti_rx_std_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_std_sparemap, m, segs, &nsegs, 0);
if (error != 0) {
m_freem(m);
return (error);
}
KASSERT(nsegs == 1, ("%s: %d segments returned!", __func__, nsegs));
if (sc->ti_cdata.ti_rx_std_chain[i] != NULL) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_std_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_std_maps[i], BUS_DMASYNC_POSTREAD);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_rx_std_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_std_maps[i]);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
map = sc->ti_cdata.ti_rx_std_maps[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_std_maps[i] = sc->ti_cdata.ti_rx_std_sparemap;
sc->ti_cdata.ti_rx_std_sparemap = map;
sc->ti_cdata.ti_rx_std_chain[i] = m;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
r = &sc->ti_rdata.ti_rx_std_ring[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_hostaddr64(&r->ti_addr, segs[0].ds_addr);
r->ti_len = segs[0].ds_len;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
r->ti_type = TI_BDTYPE_RECV_BD;
r->ti_flags = 0;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
r->ti_vlan_tag = 0;
r->ti_tcp_udp_cksum = 0;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
r->ti_flags |= TI_BDFLAG_TCP_UDP_CKSUM | TI_BDFLAG_IP_CKSUM;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
r->ti_idx = i;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_std_tag,
sc->ti_cdata.ti_rx_std_maps[i], BUS_DMASYNC_PREREAD);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Intialize a mini receive ring descriptor. This only applies to
* the Tigon 2.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_newbuf_mini(struct ti_softc *sc, int i)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
bus_dmamap_t map;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
bus_dma_segment_t segs[1];
struct mbuf *m;
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_rx_desc *r;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
int error, nsegs;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
MGETHDR(m, M_NOWAIT, MT_DATA);
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
if (m == NULL)
return (ENOBUFS);
m->m_len = m->m_pkthdr.len = MHLEN;
m_adj(m, ETHER_ALIGN);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_load_mbuf_sg(sc->ti_cdata.ti_rx_mini_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_mini_sparemap, m, segs, &nsegs, 0);
if (error != 0) {
m_freem(m);
return (error);
}
KASSERT(nsegs == 1, ("%s: %d segments returned!", __func__, nsegs));
if (sc->ti_cdata.ti_rx_mini_chain[i] != NULL) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_mini_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_mini_maps[i], BUS_DMASYNC_POSTREAD);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_rx_mini_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_mini_maps[i]);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Clean up the buffer allocation code a bit. Make sure to initialize certain critical mbuf fields to sane values. Simplify the use of ETHER_ALIGN to enforce payload alignment, and turn it on on the x86 as well as alpha since it helps with NFS which wants the payload to be longword aligned even though the hardware doesn't require it. This fixes a problem with the ti driver causing an unaligned access trap on the Alpha due to m_adj() sometimes not setting the alignment correctly because of incomplete mbuf initialization.
1999-07-23 18:46:24 +00:00
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
map = sc->ti_cdata.ti_rx_mini_maps[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_mini_maps[i] = sc->ti_cdata.ti_rx_mini_sparemap;
sc->ti_cdata.ti_rx_mini_sparemap = map;
sc->ti_cdata.ti_rx_mini_chain[i] = m;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
r = &sc->ti_rdata.ti_rx_mini_ring[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_hostaddr64(&r->ti_addr, segs[0].ds_addr);
r->ti_len = segs[0].ds_len;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
r->ti_type = TI_BDTYPE_RECV_BD;
r->ti_flags = TI_BDFLAG_MINI_RING;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
r->ti_vlan_tag = 0;
r->ti_tcp_udp_cksum = 0;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
r->ti_flags |= TI_BDFLAG_TCP_UDP_CKSUM | TI_BDFLAG_IP_CKSUM;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
r->ti_idx = i;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_mini_tag,
sc->ti_cdata.ti_rx_mini_maps[i], BUS_DMASYNC_PREREAD);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
#ifndef TI_SF_BUF_JUMBO
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Initialize a jumbo receive ring descriptor. This allocates
* a jumbo buffer from the pool managed internally by the driver.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_newbuf_jumbo(struct ti_softc *sc, int i, struct mbuf *dummy)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
bus_dmamap_t map;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
bus_dma_segment_t segs[1];
struct mbuf *m;
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_rx_desc *r;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
int error, nsegs;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
(void)dummy;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUM9BYTES);
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
if (m == NULL)
return (ENOBUFS);
m->m_len = m->m_pkthdr.len = MJUM9BYTES;
m_adj(m, ETHER_ALIGN);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_load_mbuf_sg(sc->ti_cdata.ti_rx_jumbo_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_jumbo_sparemap, m, segs, &nsegs, 0);
if (error != 0) {
m_freem(m);
return (error);
}
KASSERT(nsegs == 1, ("%s: %d segments returned!", __func__, nsegs));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
if (sc->ti_cdata.ti_rx_jumbo_chain[i] != NULL) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_jumbo_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_jumbo_maps[i], BUS_DMASYNC_POSTREAD);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_rx_jumbo_tag,
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_jumbo_maps[i]);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
map = sc->ti_cdata.ti_rx_jumbo_maps[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_cdata.ti_rx_jumbo_maps[i] = sc->ti_cdata.ti_rx_jumbo_sparemap;
sc->ti_cdata.ti_rx_jumbo_sparemap = map;
sc->ti_cdata.ti_rx_jumbo_chain[i] = m;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
r = &sc->ti_rdata.ti_rx_jumbo_ring[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_hostaddr64(&r->ti_addr, segs[0].ds_addr);
r->ti_len = segs[0].ds_len;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
r->ti_type = TI_BDTYPE_RECV_JUMBO_BD;
r->ti_flags = TI_BDFLAG_JUMBO_RING;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
r->ti_vlan_tag = 0;
r->ti_tcp_udp_cksum = 0;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
r->ti_flags |= TI_BDFLAG_TCP_UDP_CKSUM | TI_BDFLAG_IP_CKSUM;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
r->ti_idx = i;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_jumbo_tag,
sc->ti_cdata.ti_rx_jumbo_maps[i], BUS_DMASYNC_PREREAD);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#else
#if (PAGE_SIZE == 4096)
#define NPAYLOAD 2
#else
#define NPAYLOAD 1
Whitespace nits
2004-07-05 22:35:18 +00:00
#endif
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#define TCP_HDR_LEN (52 + sizeof(struct ether_header))
#define UDP_HDR_LEN (28 + sizeof(struct ether_header))
#define NFS_HDR_LEN (UDP_HDR_LEN)
style. No functional changes.
2011-11-04 18:28:10 +00:00
static int HDR_LEN = TCP_HDR_LEN;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Whitespace pass.
2004-07-05 22:46:28 +00:00
/*
* Initialize a jumbo receive ring descriptor. This allocates
* a jumbo buffer from the pool managed internally by the driver.
*/
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_newbuf_jumbo(struct ti_softc *sc, int idx, struct mbuf *m_old)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
bus_dmamap_t map;
struct mbuf *cur, *m_new = NULL;
struct mbuf *m[3] = {NULL, NULL, NULL};
struct ti_rx_desc_ext *r;
vm_page_t frame;
/* 1 extra buf to make nobufs easy*/
struct sf_buf *sf[3] = {NULL, NULL, NULL};
int i;
bus_dma_segment_t segs[4];
int nsegs;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (m_old != NULL) {
m_new = m_old;
cur = m_old->m_next;
for (i = 0; i <= NPAYLOAD; i++){
m[i] = cur;
cur = cur->m_next;
}
} else {
/* Allocate the mbufs. */
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
MGETHDR(m_new, M_NOWAIT, MT_DATA);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (m_new == NULL) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "mbuf allocation failed "
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"-- packet dropped!\n");
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
goto nobufs;
}
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
MGET(m[NPAYLOAD], M_NOWAIT, MT_DATA);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (m[NPAYLOAD] == NULL) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "cluster mbuf allocation "
"failed -- packet dropped!\n");
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
goto nobufs;
}
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
MCLGET(m[NPAYLOAD], M_NOWAIT);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if ((m[NPAYLOAD]->m_flags & M_EXT) == 0) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "mbuf allocation failed "
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"-- packet dropped!\n");
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
goto nobufs;
}
m[NPAYLOAD]->m_len = MCLBYTES;
for (i = 0; i < NPAYLOAD; i++){
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
MGET(m[i], M_NOWAIT, MT_DATA);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (m[i] == NULL) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "mbuf allocation "
"failed -- packet dropped!\n");
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
goto nobufs;
}
Eliminate vestiges of page coloring.
2011-12-15 05:07:16 +00:00
frame = vm_page_alloc(NULL, 0,
Update the Tigon 1 and 2 driver to use the sf_buf API for implementing zero-copy receive of jumbo frames. This eliminates the need for the jumbo frame allocator implemented in kern/uipc_jumbo.c and sys/jumbo.h. Remove it. Note: Zero-copy receive of jumbo frames did not work without these changes; I believe there was insufficient locking on the jumbo vm object. Tested by: ken@ Discussed with: gallatin@
2004-12-06 00:43:40 +00:00
VM_ALLOC_INTERRUPT | VM_ALLOC_NOOBJ |
VM_ALLOC_WIRED);
if (frame == NULL) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "buffer allocation "
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"failed -- packet dropped!\n");
Update the Tigon 1 and 2 driver to use the sf_buf API for implementing zero-copy receive of jumbo frames. This eliminates the need for the jumbo frame allocator implemented in kern/uipc_jumbo.c and sys/jumbo.h. Remove it. Note: Zero-copy receive of jumbo frames did not work without these changes; I believe there was insufficient locking on the jumbo vm object. Tested by: ken@ Discussed with: gallatin@
2004-12-06 00:43:40 +00:00
printf(" index %d page %d\n", idx, i);
goto nobufs;
}
sf[i] = sf_buf_alloc(frame, SFB_NOWAIT);
if (sf[i] == NULL) {
vm_page_unwire(frame, 0);
vm_page_free(frame);
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "buffer allocation "
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"failed -- packet dropped!\n");
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
printf(" index %d page %d\n", idx, i);
Whitespace pass.
2004-07-05 22:46:28 +00:00
goto nobufs;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
}
for (i = 0; i < NPAYLOAD; i++){
Whitespace pass.
2004-07-05 22:46:28 +00:00
/* Attach the buffer to the mbuf. */
Update the Tigon 1 and 2 driver to use the sf_buf API for implementing zero-copy receive of jumbo frames. This eliminates the need for the jumbo frame allocator implemented in kern/uipc_jumbo.c and sys/jumbo.h. Remove it. Note: Zero-copy receive of jumbo frames did not work without these changes; I believe there was insufficient locking on the jumbo vm object. Tested by: ken@ Discussed with: gallatin@
2004-12-06 00:43:40 +00:00
m[i]->m_data = (void *)sf_buf_kva(sf[i]);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
m[i]->m_len = PAGE_SIZE;
Update the Tigon 1 and 2 driver to use the sf_buf API for implementing zero-copy receive of jumbo frames. This eliminates the need for the jumbo frame allocator implemented in kern/uipc_jumbo.c and sys/jumbo.h. Remove it. Note: Zero-copy receive of jumbo frames did not work without these changes; I believe there was insufficient locking on the jumbo vm object. Tested by: ken@ Discussed with: gallatin@
2004-12-06 00:43:40 +00:00
MEXTADD(m[i], sf_buf_kva(sf[i]), PAGE_SIZE,
Give MEXTADD() another argument to make both void pointers to the free function controlable, instead of passing the KVA of the buffer storage as the first argument. Fix all conventional users of the API to pass the KVA of the buffer as the first argument, to make this a no-op commit. Likely break the only non-convetional user of the API, after informing the relevant committer. Update the mbuf(9) manual page, which was already out of sync on this point. Bump __FreeBSD_version to 800016 as there is no way to tell how many arguments a CPP macro needs any other way. This paves the way for giving sendfile(9) a way to wait for the passed storage to have been accessed before returning. This does not affect the memory layout or size of mbufs. Parental oversight by: sam and rwatson. No MFC is anticipated.
2008-02-01 19:36:27 +00:00
sf_buf_mext, (void*)sf_buf_kva(sf[i]), sf[i],
0, EXT_DISPOSABLE);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
m[i]->m_next = m[i+1];
}
/* link the buffers to the header */
m_new->m_next = m[0];
m_new->m_data += ETHER_ALIGN;
if (sc->ti_hdrsplit)
m_new->m_len = MHLEN - ETHER_ALIGN;
else
Whitespace pass.
2004-07-05 22:46:28 +00:00
m_new->m_len = HDR_LEN;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
m_new->m_pkthdr.len = NPAYLOAD * PAGE_SIZE + m_new->m_len;
}
/* Set up the descriptor. */
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
r = &sc->ti_rdata.ti_rx_jumbo_ring[idx];
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
sc->ti_cdata.ti_rx_jumbo_chain[idx] = m_new;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
map = sc->ti_cdata.ti_rx_jumbo_maps[i];
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (bus_dmamap_load_mbuf_sg(sc->ti_cdata.ti_rx_jumbo_tag, map, m_new,
segs, &nsegs, 0))
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
return (ENOBUFS);
if ((nsegs < 1) || (nsegs > 4))
return (ENOBUFS);
ti_hostaddr64(&r->ti_addr0, segs[0].ds_addr);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
r->ti_len0 = m_new->m_len;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
ti_hostaddr64(&r->ti_addr1, segs[1].ds_addr);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
r->ti_len1 = PAGE_SIZE;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
ti_hostaddr64(&r->ti_addr2, segs[2].ds_addr);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
r->ti_len2 = m[1]->m_ext.ext_size; /* could be PAGE_SIZE or MCLBYTES */
if (PAGE_SIZE == 4096) {
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
ti_hostaddr64(&r->ti_addr3, segs[3].ds_addr);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
r->ti_len3 = MCLBYTES;
} else {
r->ti_len3 = 0;
}
Whitespace pass.
2004-07-05 22:46:28 +00:00
r->ti_type = TI_BDTYPE_RECV_JUMBO_BD;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Whitespace pass.
2004-07-05 22:46:28 +00:00
r->ti_flags = TI_BDFLAG_JUMBO_RING|TI_RCB_FLAG_USE_EXT_RX_BD;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
r->ti_flags |= TI_BDFLAG_TCP_UDP_CKSUM|TI_BDFLAG_IP_CKSUM;
Whitespace pass.
2004-07-05 22:46:28 +00:00
r->ti_idx = idx;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_jumbo_tag, map, BUS_DMASYNC_PREREAD);
Whitespace pass.
2004-07-05 22:46:28 +00:00
return (0);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Whitespace pass.
2004-07-05 22:46:28 +00:00
nobufs:
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* Warning! :
* This can only be called before the mbufs are strung together.
Whitespace nits
2004-07-05 22:35:18 +00:00
* If the mbufs are strung together, m_freem() will free the chain,
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
* so that the later mbufs will be freed multiple times.
*/
Whitespace pass.
2004-07-05 22:46:28 +00:00
if (m_new)
m_freem(m_new);
for (i = 0; i < 3; i++) {
if (m[i])
m_freem(m[i]);
Update the Tigon 1 and 2 driver to use the sf_buf API for implementing zero-copy receive of jumbo frames. This eliminates the need for the jumbo frame allocator implemented in kern/uipc_jumbo.c and sys/jumbo.h. Remove it. Note: Zero-copy receive of jumbo frames did not work without these changes; I believe there was insufficient locking on the jumbo vm object. Tested by: ken@ Discussed with: gallatin@
2004-12-06 00:43:40 +00:00
if (sf[i])
sf_buf_mext((void *)sf_buf_kva(sf[i]), sf[i]);
Whitespace pass.
2004-07-05 22:46:28 +00:00
}
return (ENOBUFS);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
#endif
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* The standard receive ring has 512 entries in it. At 2K per mbuf cluster,
* that's 1MB or memory, which is a lot. For now, we fill only the first
* 256 ring entries and hope that our CPU is fast enough to keep up with
* the NIC.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_init_rx_ring_std(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int i;
struct ti_cmd_desc cmd;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
for (i = 0; i < TI_STD_RX_RING_CNT; i++) {
if (ti_newbuf_std(sc, i) != 0)
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENOBUFS);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
};
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_std = TI_STD_RX_RING_CNT - 1;
TI_UPDATE_STDPROD(sc, TI_STD_RX_RING_CNT - 1);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_free_rx_ring_std(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
bus_dmamap_t map;
int i;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
for (i = 0; i < TI_STD_RX_RING_CNT; i++) {
if (sc->ti_cdata.ti_rx_std_chain[i] != NULL) {
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
map = sc->ti_cdata.ti_rx_std_maps[i];
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_std_tag, map,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
BUS_DMASYNC_POSTREAD);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_rx_std_tag, map);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
m_freem(sc->ti_cdata.ti_rx_std_chain[i]);
sc->ti_cdata.ti_rx_std_chain[i] = NULL;
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bzero(sc->ti_rdata.ti_rx_std_ring, TI_STD_RX_RING_SZ);
bus_dmamap_sync(sc->ti_cdata.ti_rx_std_ring_tag,
sc->ti_cdata.ti_rx_std_ring_map, BUS_DMASYNC_PREWRITE);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_init_rx_ring_jumbo(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_cmd_desc cmd;
int i;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
for (i = 0; i < TI_JUMBO_RX_RING_CNT; i++) {
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
if (ti_newbuf_jumbo(sc, i, NULL) != 0)
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENOBUFS);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
};
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_jumbo = TI_JUMBO_RX_RING_CNT - 1;
TI_UPDATE_JUMBOPROD(sc, TI_JUMBO_RX_RING_CNT - 1);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_free_rx_ring_jumbo(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
bus_dmamap_t map;
int i;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
for (i = 0; i < TI_JUMBO_RX_RING_CNT; i++) {
if (sc->ti_cdata.ti_rx_jumbo_chain[i] != NULL) {
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
map = sc->ti_cdata.ti_rx_jumbo_maps[i];
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_jumbo_tag, map,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
BUS_DMASYNC_POSTREAD);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_rx_jumbo_tag, map);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
m_freem(sc->ti_cdata.ti_rx_jumbo_chain[i]);
sc->ti_cdata.ti_rx_jumbo_chain[i] = NULL;
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bzero(sc->ti_rdata.ti_rx_jumbo_ring, TI_JUMBO_RX_RING_SZ);
bus_dmamap_sync(sc->ti_cdata.ti_rx_jumbo_ring_tag,
sc->ti_cdata.ti_rx_jumbo_ring_map, BUS_DMASYNC_PREWRITE);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_init_rx_ring_mini(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
int i;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
for (i = 0; i < TI_MINI_RX_RING_CNT; i++) {
if (ti_newbuf_mini(sc, i) != 0)
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENOBUFS);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
};
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
sc->ti_mini = TI_MINI_RX_RING_CNT - 1;
TI_UPDATE_MINIPROD(sc, TI_MINI_RX_RING_CNT - 1);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_free_rx_ring_mini(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
bus_dmamap_t map;
int i;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_rdata.ti_rx_mini_ring == NULL)
return;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
for (i = 0; i < TI_MINI_RX_RING_CNT; i++) {
if (sc->ti_cdata.ti_rx_mini_chain[i] != NULL) {
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
map = sc->ti_cdata.ti_rx_mini_maps[i];
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_mini_tag, map,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
BUS_DMASYNC_POSTREAD);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_rx_mini_tag, map);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
m_freem(sc->ti_cdata.ti_rx_mini_chain[i]);
sc->ti_cdata.ti_rx_mini_chain[i] = NULL;
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bzero(sc->ti_rdata.ti_rx_mini_ring, TI_MINI_RX_RING_SZ);
bus_dmamap_sync(sc->ti_cdata.ti_rx_mini_ring_tag,
sc->ti_cdata.ti_rx_mini_ring_map, BUS_DMASYNC_PREWRITE);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_free_tx_ring(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_txdesc *txd;
int i;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_rdata.ti_tx_ring == NULL)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
return;
for (i = 0; i < TI_TX_RING_CNT; i++) {
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
txd = &sc->ti_cdata.ti_txdesc[i];
if (txd->tx_m != NULL) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_tx_tag, txd->tx_dmamap,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
BUS_DMASYNC_POSTWRITE);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_tx_tag,
txd->tx_dmamap);
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
m_freem(txd->tx_m);
txd->tx_m = NULL;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bzero(sc->ti_rdata.ti_tx_ring, TI_TX_RING_SZ);
bus_dmamap_sync(sc->ti_cdata.ti_tx_ring_tag,
sc->ti_cdata.ti_tx_ring_map, BUS_DMASYNC_PREWRITE);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_init_tx_ring(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_txdesc *txd;
int i;
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
STAILQ_INIT(&sc->ti_cdata.ti_txfreeq);
STAILQ_INIT(&sc->ti_cdata.ti_txbusyq);
for (i = 0; i < TI_TX_RING_CNT; i++) {
txd = &sc->ti_cdata.ti_txdesc[i];
STAILQ_INSERT_TAIL(&sc->ti_cdata.ti_txfreeq, txd, tx_q);
}
Add a transmit descriptor usage counter and use it to absolutely, positively not let ti_encap() fill up the TX ring all the way and wrap around. This fixes a potential transmit lockup where a really fast machine (or particular TX traffic pattern) can overrun the end of the ring. Reported by: John Plevyak <jplevyak@inktomi.com>
1999-06-19 00:36:56 +00:00
sc->ti_txcnt = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
sc->ti_tx_saved_considx = 0;
Cache the tx producer index instead of reading it every time ti_start is called.
2005-12-28 08:36:32 +00:00
sc->ti_tx_saved_prodidx = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_MB_SENDPROD_IDX, 0);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* The Tigon 2 firmware has a new way to add/delete multicast addresses,
* but we have to support the old way too so that Tigon 1 cards will
* work.
*/
Be consistent about functions being static. Spotted by: FlexeLint.
2002-10-16 09:04:52 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_add_mcast(struct ti_softc *sc, struct ether_addr *addr)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_cmd_desc cmd;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint16_t *m;
uint32_t ext[2] = {0, 0};
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
m = (uint16_t *)&addr->octet[0];
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
switch (sc->ti_hwrev) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
case TI_HWREV_TIGON:
CSR_WRITE_4(sc, TI_GCR_MAR0, htons(m[0]));
CSR_WRITE_4(sc, TI_GCR_MAR1, (htons(m[1]) << 16) | htons(m[2]));
TI_DO_CMD(TI_CMD_ADD_MCAST_ADDR, 0, 0);
break;
case TI_HWREV_TIGON_II:
ext[0] = htons(m[0]);
ext[1] = (htons(m[1]) << 16) | htons(m[2]);
TI_DO_CMD_EXT(TI_CMD_EXT_ADD_MCAST, 0, 0, (caddr_t)&ext, 2);
break;
default:
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "unknown hwrev\n");
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
}
}
Be consistent about functions being static. Spotted by: FlexeLint.
2002-10-16 09:04:52 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_del_mcast(struct ti_softc *sc, struct ether_addr *addr)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_cmd_desc cmd;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint16_t *m;
uint32_t ext[2] = {0, 0};
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
m = (uint16_t *)&addr->octet[0];
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
switch (sc->ti_hwrev) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
case TI_HWREV_TIGON:
CSR_WRITE_4(sc, TI_GCR_MAR0, htons(m[0]));
CSR_WRITE_4(sc, TI_GCR_MAR1, (htons(m[1]) << 16) | htons(m[2]));
TI_DO_CMD(TI_CMD_DEL_MCAST_ADDR, 0, 0);
break;
case TI_HWREV_TIGON_II:
ext[0] = htons(m[0]);
ext[1] = (htons(m[1]) << 16) | htons(m[2]);
TI_DO_CMD_EXT(TI_CMD_EXT_DEL_MCAST, 0, 0, (caddr_t)&ext, 2);
break;
default:
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "unknown hwrev\n");
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
}
}
/*
* Configure the Tigon's multicast address filter.
*
* The actual multicast table management is a bit of a pain, thanks to
* slight brain damage on the part of both Alteon and us. With our
* multicast code, we are only alerted when the multicast address table
* changes and at that point we only have the current list of addresses:
* we only know the current state, not the previous state, so we don't
* actually know what addresses were removed or added. The firmware has
* state, but we can't get our grubby mits on it, and there is no 'delete
* all multicast addresses' command. Hence, we have to maintain our own
* state so we know what addresses have been programmed into the NIC at
* any given time.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_setmulti(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ifnet *ifp;
struct ifmultiaddr *ifma;
struct ti_cmd_desc cmd;
struct ti_mc_entry *mc;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t intrs;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (ifp->if_flags & IFF_ALLMULTI) {
TI_DO_CMD(TI_CMD_SET_ALLMULTI, TI_CMD_CODE_ALLMULTI_ENB, 0);
return;
} else {
TI_DO_CMD(TI_CMD_SET_ALLMULTI, TI_CMD_CODE_ALLMULTI_DIS, 0);
}
/* Disable interrupts. */
intrs = CSR_READ_4(sc, TI_MB_HOSTINTR);
CSR_WRITE_4(sc, TI_MB_HOSTINTR, 1);
/* First, zot all the existing filters. */
Mechanical change to use <sys/queue.h> macro API instead of fondling implementation details. Created with: sed(1) Reviewed by: md5(1)
2001-02-04 13:13:25 +00:00
while (SLIST_FIRST(&sc->ti_mc_listhead) != NULL) {
mc = SLIST_FIRST(&sc->ti_mc_listhead);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ti_del_mcast(sc, &mc->mc_addr);
SLIST_REMOVE_HEAD(&sc->ti_mc_listhead, mc_entries);
free(mc, M_DEVBUF);
}
/* Now program new ones. */
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_rlock(ifp);
Convert if_multiaddrs from LIST to TAILQ so that it can be traversed backwards in the three drivers which want to do that. Reviewed by: mikeh
2001-02-06 10:12:15 +00:00
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
mc = malloc(sizeof(struct ti_mc_entry), M_DEVBUF, M_NOWAIT);
deal with malloc failure when setting up the multicast filter Noticed by: Coverity Prevent analysis tool
2005-03-26 23:26:49 +00:00
if (mc == NULL) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev,
"no memory for mcast filter entry\n");
deal with malloc failure when setting up the multicast filter Noticed by: Coverity Prevent analysis tool
2005-03-26 23:26:49 +00:00
continue;
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
(char *)&mc->mc_addr, ETHER_ADDR_LEN);
SLIST_INSERT_HEAD(&sc->ti_mc_listhead, mc, mc_entries);
ti_add_mcast(sc, &mc->mc_addr);
}
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_runlock(ifp);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Re-enable interrupts. */
CSR_WRITE_4(sc, TI_MB_HOSTINTR, intrs);
}
/*
* Check to see if the BIOS has configured us for a 64 bit slot when
* we aren't actually in one. If we detect this condition, we can work
* around it on the Tigon 2 by setting a bit in the PCI state register,
* but for the Tigon 1 we must give up and abort the interface attach.
*/
style. No functional changes.
2011-11-10 22:15:11 +00:00
static int
ti_64bitslot_war(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (!(CSR_READ_4(sc, TI_PCI_STATE) & TI_PCISTATE_32BIT_BUS)) {
CSR_WRITE_4(sc, 0x600, 0);
CSR_WRITE_4(sc, 0x604, 0);
CSR_WRITE_4(sc, 0x600, 0x5555AAAA);
if (CSR_READ_4(sc, 0x604) == 0x5555AAAA) {
if (sc->ti_hwrev == TI_HWREV_TIGON)
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (EINVAL);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
else {
TI_SETBIT(sc, TI_PCI_STATE,
TI_PCISTATE_32BIT_BUS);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
}
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Do endian, PCI and DMA initialization. Also check the on-board ROM
* self-test results.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_chipinit(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t cacheline;
uint32_t pci_writemax = 0;
uint32_t hdrsplit;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Initialize link to down state. */
sc->ti_linkstat = TI_EV_CODE_LINK_DOWN;
/* Set endianness before we access any non-PCI registers. */
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
#if 0 && BYTE_ORDER == BIG_ENDIAN
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_MISC_HOST_CTL,
TI_MHC_BIGENDIAN_INIT | (TI_MHC_BIGENDIAN_INIT << 24));
#else
CSR_WRITE_4(sc, TI_MISC_HOST_CTL,
TI_MHC_LITTLEENDIAN_INIT | (TI_MHC_LITTLEENDIAN_INIT << 24));
#endif
/* Check the ROM failed bit to see if self-tests passed. */
if (CSR_READ_4(sc, TI_CPU_STATE) & TI_CPUSTATE_ROMFAIL) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "board self-diagnostics failed!\n");
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENODEV);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/* Halt the CPU. */
TI_SETBIT(sc, TI_CPU_STATE, TI_CPUSTATE_HALT);
/* Figure out the hardware revision. */
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
switch (CSR_READ_4(sc, TI_MISC_HOST_CTL) & TI_MHC_CHIP_REV_MASK) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
case TI_REV_TIGON_I:
sc->ti_hwrev = TI_HWREV_TIGON;
break;
case TI_REV_TIGON_II:
sc->ti_hwrev = TI_HWREV_TIGON_II;
break;
default:
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "unsupported chip revision\n");
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENODEV);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/* Do special setup for Tigon 2. */
if (sc->ti_hwrev == TI_HWREV_TIGON_II) {
TI_SETBIT(sc, TI_CPU_CTL_B, TI_CPUSTATE_HALT);
Fix the definitions for memory bank sizes, which I somehow got wrong. The constant I was using was correct, but I mislabeled it as 256K when it should have been 512K. This doesn't actually change the code, but it clarifies things somewhat. Submitted by: Chuck Cranor <chuck@research.att.com>
2001-04-26 16:40:45 +00:00
TI_SETBIT(sc, TI_MISC_LOCAL_CTL, TI_MLC_SRAM_BANK_512K);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_SETBIT(sc, TI_MISC_CONF, TI_MCR_SRAM_SYNCHRONOUS);
}
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* We don't have firmware source for the Tigon 1, so Tigon 1 boards
* can't do header splitting.
*/
#ifdef TI_JUMBO_HDRSPLIT
if (sc->ti_hwrev != TI_HWREV_TIGON)
sc->ti_hdrsplit = 1;
else
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev,
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"can't do header splitting on a Tigon I board\n");
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#endif /* TI_JUMBO_HDRSPLIT */
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Set up the PCI state register. */
CSR_WRITE_4(sc, TI_PCI_STATE, TI_PCI_READ_CMD|TI_PCI_WRITE_CMD);
if (sc->ti_hwrev == TI_HWREV_TIGON_II) {
TI_SETBIT(sc, TI_PCI_STATE, TI_PCISTATE_USE_MEM_RD_MULT);
}
/* Clear the read/write max DMA parameters. */
TI_CLRBIT(sc, TI_PCI_STATE, (TI_PCISTATE_WRITE_MAXDMA|
TI_PCISTATE_READ_MAXDMA));
/* Get cache line size. */
cacheline = CSR_READ_4(sc, TI_PCI_BIST) & 0xFF;
/*
* If the system has set enabled the PCI memory write
* and invalidate command in the command register, set
* the write max parameter accordingly. This is necessary
* to use MWI with the Tigon 2.
*/
if (CSR_READ_4(sc, TI_PCI_CMDSTAT) & PCIM_CMD_MWIEN) {
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
switch (cacheline) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
case 1:
case 4:
case 8:
case 16:
case 32:
case 64:
break;
default:
/* Disable PCI memory write and invalidate. */
if (bootverbose)
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "cache line size %d"
" not supported; disabling PCI MWI\n",
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
cacheline);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_PCI_CMDSTAT, CSR_READ_4(sc,
TI_PCI_CMDSTAT) & ~PCIM_CMD_MWIEN);
break;
}
}
TI_SETBIT(sc, TI_PCI_STATE, pci_writemax);
/* This sets the min dma param all the way up (0xff). */
TI_SETBIT(sc, TI_PCI_STATE, TI_PCISTATE_MINDMA);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (sc->ti_hdrsplit)
hdrsplit = TI_OPMODE_JUMBO_HDRSPLIT;
else
hdrsplit = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Configure DMA variables. */
#if BYTE_ORDER == BIG_ENDIAN
CSR_WRITE_4(sc, TI_GCR_OPMODE, TI_OPMODE_BYTESWAP_BD |
TI_OPMODE_BYTESWAP_DATA | TI_OPMODE_WORDSWAP_BD |
TI_OPMODE_WARN_ENB | TI_OPMODE_FATAL_ENB |
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
TI_OPMODE_DONT_FRAG_JUMBO | hdrsplit);
#else /* BYTE_ORDER */
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_OPMODE, TI_OPMODE_BYTESWAP_DATA|
TI_OPMODE_WORDSWAP_BD|TI_OPMODE_DONT_FRAG_JUMBO|
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
TI_OPMODE_WARN_ENB|TI_OPMODE_FATAL_ENB | hdrsplit);
#endif /* BYTE_ORDER */
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Only allow 1 DMA channel to be active at a time.
* I don't think this is a good idea, but without it
* the firmware racks up lots of nicDmaReadRingFull
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
* errors. This is not compatible with hardware checksums.
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
*/
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if ((sc->ti_ifp->if_capenable & (IFCAP_TXCSUM | IFCAP_RXCSUM)) == 0)
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
TI_SETBIT(sc, TI_GCR_OPMODE, TI_OPMODE_1_DMA_ACTIVE);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Recommended settings from Tigon manual. */
CSR_WRITE_4(sc, TI_GCR_DMA_WRITECFG, TI_DMA_STATE_THRESH_8W);
CSR_WRITE_4(sc, TI_GCR_DMA_READCFG, TI_DMA_STATE_THRESH_8W);
if (ti_64bitslot_war(sc)) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "bios thinks we're in a 64 bit slot, "
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
"but we aren't");
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (EINVAL);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Initialize the general information block and firmware, and
* start the CPU(s) running.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_gibinit(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ifnet *ifp;
struct ti_rcb *rcb;
int i;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Disable interrupts for now. */
CSR_WRITE_4(sc, TI_MB_HOSTINTR, 1);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Tell the chip where to find the general information block. */
CSR_WRITE_4(sc, TI_GCR_GENINFO_HI,
(uint64_t)sc->ti_rdata.ti_info_paddr >> 32);
CSR_WRITE_4(sc, TI_GCR_GENINFO_LO,
sc->ti_rdata.ti_info_paddr & 0xFFFFFFFF);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Load the firmware into SRAM. */
ti_loadfw(sc);
/* Set up the contents of the general info and ring control blocks. */
/* Set up the event ring and producer pointer. */
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bzero(sc->ti_rdata.ti_event_ring, TI_EVENT_RING_SZ);
rcb = &sc->ti_rdata.ti_info->ti_ev_rcb;
ti_hostaddr64(&rcb->ti_hostaddr, sc->ti_rdata.ti_event_ring_paddr);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
rcb->ti_flags = 0;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
ti_hostaddr64(&sc->ti_rdata.ti_info->ti_ev_prodidx_ptr,
sc->ti_rdata.ti_status_paddr +
offsetof(struct ti_status, ti_ev_prodidx_r));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
sc->ti_ev_prodidx.ti_idx = 0;
CSR_WRITE_4(sc, TI_GCR_EVENTCONS_IDX, 0);
sc->ti_ev_saved_considx = 0;
/* Set up the command ring and producer mailbox. */
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
rcb = &sc->ti_rdata.ti_info->ti_cmd_rcb;
ti_hostaddr64(&rcb->ti_hostaddr, TI_GCR_NIC_ADDR(TI_GCR_CMDRING));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
rcb->ti_flags = 0;
rcb->ti_max_len = 0;
for (i = 0; i < TI_CMD_RING_CNT; i++) {
CSR_WRITE_4(sc, TI_GCR_CMDRING + (i * 4), 0);
}
CSR_WRITE_4(sc, TI_GCR_CMDCONS_IDX, 0);
CSR_WRITE_4(sc, TI_MB_CMDPROD_IDX, 0);
sc->ti_cmd_saved_prodidx = 0;
/*
* Assign the address of the stats refresh buffer.
* We re-use the current stats buffer for this to
* conserve memory.
*/
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bzero(&sc->ti_rdata.ti_info->ti_stats, sizeof(struct ti_stats));
ti_hostaddr64(&sc->ti_rdata.ti_info->ti_refresh_stats_ptr,
sc->ti_rdata.ti_info_paddr + offsetof(struct ti_gib, ti_stats));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Set up the standard receive ring. */
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
rcb = &sc->ti_rdata.ti_info->ti_std_rx_rcb;
ti_hostaddr64(&rcb->ti_hostaddr, sc->ti_rdata.ti_rx_std_ring_paddr);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
rcb->ti_max_len = TI_FRAMELEN;
rcb->ti_flags = 0;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
rcb->ti_flags |= TI_RCB_FLAG_TCP_UDP_CKSUM |
TI_RCB_FLAG_IP_CKSUM | TI_RCB_FLAG_NO_PHDR_CKSUM;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
rcb->ti_flags |= TI_RCB_FLAG_VLAN_ASSIST;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Set up the jumbo receive ring. */
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
rcb = &sc->ti_rdata.ti_info->ti_jumbo_rx_rcb;
ti_hostaddr64(&rcb->ti_hostaddr, sc->ti_rdata.ti_rx_jumbo_ring_paddr);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
#ifndef TI_SF_BUF_JUMBO
rcb->ti_max_len = MJUM9BYTES - ETHER_ALIGN;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
rcb->ti_flags = 0;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#else
rcb->ti_max_len = PAGE_SIZE;
rcb->ti_flags = TI_RCB_FLAG_USE_EXT_RX_BD;
#endif
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
rcb->ti_flags |= TI_RCB_FLAG_TCP_UDP_CKSUM |
TI_RCB_FLAG_IP_CKSUM | TI_RCB_FLAG_NO_PHDR_CKSUM;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
rcb->ti_flags |= TI_RCB_FLAG_VLAN_ASSIST;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Set up the mini ring. Only activated on the
* Tigon 2 but the slot in the config block is
* still there on the Tigon 1.
*/
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
rcb = &sc->ti_rdata.ti_info->ti_mini_rx_rcb;
ti_hostaddr64(&rcb->ti_hostaddr, sc->ti_rdata.ti_rx_mini_ring_paddr);
Remember to account for ETHER_ALIGN when setting the maxmimum packet length for mini receive ring. The max length was MHLEN, however the mbufs are actually shortened to MHLEN - ETHER_ALIGN to force payload alignment. PR: 13793
1999-09-17 18:04:14 +00:00
rcb->ti_max_len = MHLEN - ETHER_ALIGN;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON)
rcb->ti_flags = TI_RCB_FLAG_RING_DISABLED;
else
rcb->ti_flags = 0;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
rcb->ti_flags |= TI_RCB_FLAG_TCP_UDP_CKSUM |
TI_RCB_FLAG_IP_CKSUM | TI_RCB_FLAG_NO_PHDR_CKSUM;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
rcb->ti_flags |= TI_RCB_FLAG_VLAN_ASSIST;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Set up the receive return ring.
*/
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
rcb = &sc->ti_rdata.ti_info->ti_return_rcb;
ti_hostaddr64(&rcb->ti_hostaddr, sc->ti_rdata.ti_rx_return_ring_paddr);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
rcb->ti_flags = 0;
rcb->ti_max_len = TI_RETURN_RING_CNT;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
ti_hostaddr64(&sc->ti_rdata.ti_info->ti_return_prodidx_ptr,
sc->ti_rdata.ti_status_paddr +
offsetof(struct ti_status, ti_return_prodidx_r));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Set up the tx ring. Note: for the Tigon 2, we have the option
* of putting the transmit ring in the host's address space and
* letting the chip DMA it instead of leaving the ring in the NIC's
* memory and accessing it through the shared memory region. We
* do this for the Tigon 2, but it doesn't work on the Tigon 1,
* so we have to revert to the shared memory scheme if we detect
* a Tigon 1 chip.
*/
CSR_WRITE_4(sc, TI_WINBASE, TI_TX_RING_BASE);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_rdata.ti_tx_ring != NULL)
bzero(sc->ti_rdata.ti_tx_ring, TI_TX_RING_SZ);
rcb = &sc->ti_rdata.ti_info->ti_tx_rcb;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON)
rcb->ti_flags = 0;
else
rcb->ti_flags = TI_RCB_FLAG_HOST_RING;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (sc->ti_ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
rcb->ti_flags |= TI_RCB_FLAG_VLAN_ASSIST;
if (sc->ti_ifp->if_capenable & IFCAP_TXCSUM)
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
rcb->ti_flags |= TI_RCB_FLAG_TCP_UDP_CKSUM |
TI_RCB_FLAG_IP_CKSUM | TI_RCB_FLAG_NO_PHDR_CKSUM;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
rcb->ti_max_len = TI_TX_RING_CNT;
if (sc->ti_hwrev == TI_HWREV_TIGON)
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
ti_hostaddr64(&rcb->ti_hostaddr, TI_TX_RING_BASE);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
else
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
ti_hostaddr64(&rcb->ti_hostaddr,
sc->ti_rdata.ti_tx_ring_paddr);
ti_hostaddr64(&sc->ti_rdata.ti_info->ti_tx_considx_ptr,
sc->ti_rdata.ti_status_paddr +
offsetof(struct ti_status, ti_tx_considx_r));
bus_dmamap_sync(sc->ti_cdata.ti_gib_tag, sc->ti_cdata.ti_gib_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
bus_dmamap_sync(sc->ti_cdata.ti_status_tag, sc->ti_cdata.ti_status_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
bus_dmamap_sync(sc->ti_cdata.ti_event_ring_tag,
sc->ti_cdata.ti_event_ring_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
if (sc->ti_rdata.ti_tx_ring != NULL)
bus_dmamap_sync(sc->ti_cdata.ti_tx_ring_tag,
sc->ti_cdata.ti_tx_ring_map, BUS_DMASYNC_PREWRITE);
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
Export sysctl node for various interrupt moderation parameters and have administrators control them. ti(4) provides a character device to control various other features of driver via ioctls but users had to write their own code to manipulate these parameters. It seems some default values for these parameters are not optimal on today's system but leave it as it was and let administrators change them. The following parameters could be changed: dev.ti.%d.rx_coal_ticks dev.ti.%d.rx_max_coal_bds dev.ti.%d.tx_coal_ticks dev.ti.%d.tx_max_coal_bds dev.ti.%d.tx_buf_ratio dev.ti.%d.stat_ticks The interface has to be brought down and up again before a change takes effect. ti(4) controller supports hardware MAC counters with additional DMA statistics. So it's doable to export these counters via sysctl interface. Unfortunately, these counters are cumulative such that driver have to either send an explicit clear command to controller after extracting them or have to maintain internal counters to get actual changes. Neither look good to me so counters were not exported via sysctl.
2011-11-14 19:10:20 +00:00
/* Set up tunables */
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#if 0
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (ifp->if_mtu > ETHERMTU + ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_RX_COAL_TICKS,
(sc->ti_rx_coal_ticks / 10));
else
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#endif
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_RX_COAL_TICKS, sc->ti_rx_coal_ticks);
CSR_WRITE_4(sc, TI_GCR_TX_COAL_TICKS, sc->ti_tx_coal_ticks);
CSR_WRITE_4(sc, TI_GCR_STAT_TICKS, sc->ti_stat_ticks);
CSR_WRITE_4(sc, TI_GCR_RX_MAX_COAL_BD, sc->ti_rx_max_coal_bds);
CSR_WRITE_4(sc, TI_GCR_TX_MAX_COAL_BD, sc->ti_tx_max_coal_bds);
CSR_WRITE_4(sc, TI_GCR_TX_BUFFER_RATIO, sc->ti_tx_buf_ratio);
/* Turn interrupts on. */
CSR_WRITE_4(sc, TI_GCR_MASK_INTRS, 0);
CSR_WRITE_4(sc, TI_MB_HOSTINTR, 0);
/* Start CPU. */
TI_CLRBIT(sc, TI_CPU_STATE, (TI_CPUSTATE_HALT|TI_CPUSTATE_STEP));
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Probe for a Tigon chip. Check the PCI vendor and device IDs
* against our list and return its name if we find a match.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_probe(device_t dev)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
const struct ti_type *t;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
t = ti_devs;
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
while (t->ti_name != NULL) {
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
if ((pci_get_vendor(dev) == t->ti_vid) &&
(pci_get_device(dev) == t->ti_did)) {
device_set_desc(dev, t->ti_name);
Return BUS_PROBE_DEFAULT instead of 0.
2005-02-24 21:32:56 +00:00
return (BUS_PROBE_DEFAULT);
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
t++;
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENXIO);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_attach(device_t dev)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ifnet *ifp;
struct ti_softc *sc;
int error = 0, rid;
u_char eaddr[6];
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
sc = device_get_softc(dev);
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
sc->ti_dev = dev;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Change callers of mtx_init() to pass in an appropriate lock type name. In most cases NULL is passed, but in some cases such as network driver locks (which use the MTX_NETWORK_LOCK macro) and UMA zone locks, a name is used. Tested on: i386, alpha, sparc64
2002-04-04 21:03:38 +00:00
mtx_init(&sc->ti_mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
MTX_DEF);
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
callout_init_mtx(&sc->ti_watchdog, &sc->ti_mtx, 0);
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
ifmedia_init(&sc->ifmedia, IFM_IMASK, ti_ifmedia_upd, ti_ifmedia_sts);
Fix ifnet fallout in if_ti. Reviewed by: brooks Approved by: re
2005-07-07 01:05:45 +00:00
ifp = sc->ti_ifp = if_alloc(IFT_ETHER);
if (ifp == NULL) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
device_printf(dev, "can not if_alloc()\n");
Fix ifnet fallout in if_ti. Reviewed by: brooks Approved by: re
2005-07-07 01:05:45 +00:00
error = ENOSPC;
goto fail;
}
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
sc->ti_ifp->if_hwassist = TI_CSUM_FEATURES;
sc->ti_ifp->if_capabilities = IFCAP_TXCSUM | IFCAP_RXCSUM;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
sc->ti_ifp->if_capenable = sc->ti_ifp->if_capabilities;
Initialize/grab the mutex earlier in the attach phase, so that bailing out to the fail: label where we release/destroy the mutex will work without exploding.
2000-12-04 22:46:50 +00:00
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Map control/status registers.
*/
Big round of minor updates: - Use pci_get_powerstate()/pci_set_powerstate() in all the other drivers that need them so we don't have to fiddle with the PCI power management registers directly. - Use pci_enable_busmaster()/pci_enable_io() to turn on busmastering and PIO/memory mapped accesses. - Add support to the RealTek driver for the D-Link DFE-530TX+ which has a RealTek 8139 with its own PCI ID. (Submitted by Jason Wright) - Have the SiS 900/National DP83815 driver be sure to disable PME mode in sis_reset(). This apparently fixes a problem on some motherboards where the DP83815 chip fails to receive packets. (Submitted by Chuck McCrobie <mccrobie@cablespeed.com>)
2001-02-21 20:54:22 +00:00
pci_enable_busmaster(dev);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Remove ti_unit member variable in softc. While I'm here use PCIR_BAR macro.
2011-11-07 18:19:20 +00:00
rid = PCIR_BAR(0);
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->ti_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
Remove some remnant alpha hacks. Approved by: PCI-maintainers (imp, jhb)
2008-04-26 14:13:48 +00:00
RF_ACTIVE);
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
if (sc->ti_res == NULL) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
device_printf(dev, "couldn't map memory\n");
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
error = ENXIO;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
goto fail;
}
Grrr. Return the rman_get_bustag()/rman_get_bushandle() lines to their proper place in ti_attach(). I'm positive I typed them in there, but they must have fallen victim to a drive-by cut & pasting.
1999-07-23 16:21:43 +00:00
sc->ti_btag = rman_get_bustag(sc->ti_res);
sc->ti_bhandle = rman_get_bushandle(sc->ti_res);
Roar! Finish what I started last night: somehow only the header file change got committed.
1999-07-27 13:54:15 +00:00
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
/* Allocate interrupt */
rid = 0;
Whitespace nits
2004-07-05 22:35:18 +00:00
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->ti_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
RF_SHAREABLE | RF_ACTIVE);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
if (sc->ti_irq == NULL) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
device_printf(dev, "couldn't map interrupt\n");
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
error = ENXIO;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
goto fail;
}
if (ti_chipinit(sc)) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
device_printf(dev, "chip initialization failed\n");
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
error = ENXIO;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
goto fail;
}
/* Zero out the NIC's on-board SRAM. */
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
ti_mem_zero(sc, 0x2000, 0x100000 - 0x2000);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Init again -- zeroing memory may have clobbered some registers. */
if (ti_chipinit(sc)) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
device_printf(dev, "chip initialization failed\n");
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
error = ENXIO;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
goto fail;
}
/*
* Get station address from the EEPROM. Note: the manual states
* that the MAC address is at offset 0x8c, however the data is
* stored as two longwords (since that's how it's loaded into
Preceed/preceeding are not english words. Use precede and preceding.
2001-02-18 10:43:53 +00:00
* the NIC). This means the MAC address is actually preceded
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
* by two zero bytes. We need to skip over those.
*/
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (ti_read_eeprom(sc, eaddr, TI_EE_MAC_OFFSET + 2, ETHER_ADDR_LEN)) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
device_printf(dev, "failed to read station address\n");
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
error = ENXIO;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
goto fail;
}
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
/* Allocate working area for memory dump. */
sc->ti_membuf = malloc(sizeof(uint8_t) * TI_WINLEN, M_DEVBUF, M_NOWAIT);
sc->ti_membuf2 = malloc(sizeof(uint8_t) * TI_WINLEN, M_DEVBUF,
M_NOWAIT);
if (sc->ti_membuf == NULL || sc->ti_membuf2 == NULL) {
device_printf(dev, "cannot allocate memory buffer\n");
error = ENOMEM;
goto fail;
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if ((error = ti_dma_alloc(sc)) != 0)
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
goto fail;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
/*
* We really need a better way to tell a 1000baseTX card
* from a 1000baseSX one, since in theory there could be
* OEMed 1000baseTX cards from lame vendors who aren't
* clever enough to change the PCI ID. For the moment
* though, the AceNIC is the only copper card available.
*/
if (pci_get_vendor(dev) == ALT_VENDORID &&
pci_get_device(dev) == ALT_DEVICEID_ACENIC_COPPER)
sc->ti_copper = 1;
Add support for the Netgear GA620T copper gigabit card.
2000-08-02 18:49:17 +00:00
/* Ok, it's not the only copper card available. */
if (pci_get_vendor(dev) == NG_VENDORID &&
pci_get_device(dev) == NG_DEVICEID_GA620T)
sc->ti_copper = 1;
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
Export sysctl node for various interrupt moderation parameters and have administrators control them. ti(4) provides a character device to control various other features of driver via ioctls but users had to write their own code to manipulate these parameters. It seems some default values for these parameters are not optimal on today's system but leave it as it was and let administrators change them. The following parameters could be changed: dev.ti.%d.rx_coal_ticks dev.ti.%d.rx_max_coal_bds dev.ti.%d.tx_coal_ticks dev.ti.%d.tx_max_coal_bds dev.ti.%d.tx_buf_ratio dev.ti.%d.stat_ticks The interface has to be brought down and up again before a change takes effect. ti(4) controller supports hardware MAC counters with additional DMA statistics. So it's doable to export these counters via sysctl interface. Unfortunately, these counters are cumulative such that driver have to either send an explicit clear command to controller after extracting them or have to maintain internal counters to get actual changes. Neither look good to me so counters were not exported via sysctl.
2011-11-14 19:10:20 +00:00
/* Set default tunable values. */
ti_sysctl_node(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Set up ifnet structure */
ifp->if_softc = sc;
Replace the if_name and if_unit members of struct ifnet with new members if_xname, if_dname, and if_dunit. if_xname is the name of the interface and if_dname/unit are the driver name and instance. This change paves the way for interface renaming and enhanced pseudo device creation and configuration symantics. Approved By: re (in principle) Reviewed By: njl, imp Tested On: i386, amd64, sparc64 Obtained From: NetBSD (if_xname)
2003-10-31 18:32:15 +00:00
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifp->if_ioctl = ti_ioctl;
ifp->if_start = ti_start;
ifp->if_init = ti_init;
Announce IFCAP_LINKSTATE capability and let network stack know link state changes. Hide superfluous link up/down message under bootverbose since if_link_state_change(9) shows that information. While I'm here, change baudrate with the resolved speed of the established link instead of blindly setting it 1G. Unfortunately, it seems there is no way to differentiate 10/100Mbps from non-gigabit link so just assume we established a 100Mbps link if current link is not a gigabit link.
2011-11-04 20:43:37 +00:00
ifp->if_baudrate = IF_Gbps(1UL);
Implement altq(4) support. While I'm here fix a logic error in r227098 where it didn't re-enable interrupts when TX queue is empty.
2011-11-04 23:34:54 +00:00
ifp->if_snd.ifq_drv_maxlen = TI_TX_RING_CNT - 1;
IFQ_SET_MAXLEN(&ifp->if_snd, ifp->if_snd.ifq_drv_maxlen);
IFQ_SET_READY(&ifp->if_snd);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Set up ifmedia support. */
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
if (sc->ti_copper) {
/*
* Copper cards allow manual 10/100 mode selection,
* but not manual 1000baseTX mode selection. Why?
* Becuase currently there's no way to specify the
* master/slave setting through the firmware interface,
* so Alteon decided to just bag it and handle it
* via autonegotiation.
*/
ifmedia_add(&sc->ifmedia, IFM_ETHER|IFM_10_T, 0, NULL);
ifmedia_add(&sc->ifmedia,
IFM_ETHER|IFM_10_T|IFM_FDX, 0, NULL);
ifmedia_add(&sc->ifmedia, IFM_ETHER|IFM_100_TX, 0, NULL);
ifmedia_add(&sc->ifmedia,
IFM_ETHER|IFM_100_TX|IFM_FDX, 0, NULL);
Follow NetBSD and s/IFM_1000_TX/IFM_1000_T/
2002-04-28 20:34:20 +00:00
ifmedia_add(&sc->ifmedia, IFM_ETHER|IFM_1000_T, 0, NULL);
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
ifmedia_add(&sc->ifmedia,
Follow NetBSD and s/IFM_1000_TX/IFM_1000_T/
2002-04-28 20:34:20 +00:00
IFM_ETHER|IFM_1000_T|IFM_FDX, 0, NULL);
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
} else {
/* Fiber cards don't support 10/100 modes. */
ifmedia_add(&sc->ifmedia, IFM_ETHER|IFM_1000_SX, 0, NULL);
ifmedia_add(&sc->ifmedia,
IFM_ETHER|IFM_1000_SX|IFM_FDX, 0, NULL);
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifmedia_add(&sc->ifmedia, IFM_ETHER|IFM_AUTO, 0, NULL);
ifmedia_set(&sc->ifmedia, IFM_ETHER|IFM_AUTO);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* We're assuming here that card initialization is a sequential
* thing. If it isn't, multiple cards probing at the same time
* could stomp on the list of softcs here.
*/
/* Register the device */
Remove ti_unit member variable in softc. While I'm here use PCIR_BAR macro.
2011-11-07 18:19:20 +00:00
sc->dev = make_dev(&ti_cdevsw, device_get_unit(dev), UID_ROOT,
GID_OPERATOR, 0600, "ti%d", device_get_unit(dev));
Sanitize the code relating to the /dev/ti%d entries. In particular evict the evil vnode sniffing code and use destroy_dev() instead.
2003-10-10 20:35:28 +00:00
sc->dev->si_drv1 = sc;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
Make all Ethernet drivers attach using ether_ifattach() and detach using ether_ifdetach(). The former consolidates the operations of if_attach(), ng_ether_attach(), and bpfattach(). The latter consolidates the corresponding detach operations. Reviewed by: julian, freebsd-net
2000-07-13 22:54:34 +00:00
* Call MI attach routine.
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
*/
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ether_ifattach(ifp, eaddr);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
/* VLAN capability setup. */
ifp->if_capabilities |= IFCAP_VLAN_MTU | IFCAP_VLAN_HWCSUM |
IFCAP_VLAN_HWTAGGING;
ifp->if_capenable = ifp->if_capabilities;
/* Tell the upper layer we support VLAN over-sized frames. */
ifp->if_hdrlen = sizeof(struct ether_vlan_header);
Announce IFCAP_LINKSTATE capability and let network stack know link state changes. Hide superfluous link up/down message under bootverbose since if_link_state_change(9) shows that information. While I'm here, change baudrate with the resolved speed of the established link instead of blindly setting it 1G. Unfortunately, it seems there is no way to differentiate 10/100Mbps from non-gigabit link so just assume we established a 100Mbps link if current link is not a gigabit link.
2011-11-04 20:43:37 +00:00
/* Driver supports link state tracking. */
ifp->if_capabilities |= IFCAP_LINKSTATE;
ifp->if_capenable |= IFCAP_LINKSTATE;
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
/* Hook interrupt last to avoid having to lock softc */
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
error = bus_setup_intr(dev, sc->ti_irq, INTR_TYPE_NET|INTR_MPSAFE,
o break newbus api: add a new argument of type driver_filter_t to bus_setup_intr() o add an int return code to all fast handlers o retire INTR_FAST/IH_FAST For more info: http://docs.freebsd.org/cgi/getmsg.cgi?fetch=465712+0+current/freebsd-current Reviewed by: many Approved by: re@
2007-02-23 12:19:07 +00:00
NULL, ti_intr, sc, &sc->ti_intrhand);
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
if (error) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
device_printf(dev, "couldn't set up irq\n");
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
goto fail;
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
fail:
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
if (error)
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
ti_detach(dev);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (error);
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
}
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
/*
* Shutdown hardware and free up resources. This can be called any
* time after the mutex has been initialized. It is called in both
* the error case in attach and the normal detach case so it needs
* to be careful about only freeing resources that have actually been
* allocated.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_detach(device_t dev)
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
struct ifnet *ifp;
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
sc = device_get_softc(dev);
If resource allocation fails, we could wind up freeing the cdev without it being allocated. Add a simple check for this. Submitted by: yongari
2005-03-31 17:16:40 +00:00
if (sc->dev)
destroy_dev(sc->dev);
Fix KASSERT syntax error.
2003-04-01 08:57:28 +00:00
KASSERT(mtx_initialized(&sc->ti_mtx), ("ti mutex not initialized"));
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
if (device_is_attached(dev)) {
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
ether_ifdetach(ifp);
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
TI_LOCK(sc);
ti_stop(sc);
TI_UNLOCK(sc);
}
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
/* These should only be active if attach succeeded */
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
callout_drain(&sc->ti_watchdog);
bus_generic_detach(dev);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
ti_dma_free(sc);
Revise attach/detach resource cleanup - Unconditionally call *_stop() if device is in the tree. This is to prevent callouts from happening after the device is gone. Checks for bus_child_present() should be added in the future to keep from touching potentially non-existent hardware in *_detach(). Found by iedowse@. - Always check for and free miibus children, even if the device is not in the tree since some failure cases could have gotten here. - Call ether_ifdetach() in the irq setup failure case - ti(4), xl(4): move ifmedia_init() calls to the beginning of attach so that ifmedia_removeall() can be unconditionally called on detach. There is no way to detect whether ifmedia has been initialized without using a separate variable (as tl(4) does). - Add comments to indicate assumptions of code path
2003-04-17 20:32:06 +00:00
ifmedia_removeall(&sc->ifmedia);
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
if (sc->ti_intrhand)
bus_teardown_intr(dev, sc->ti_irq, sc->ti_intrhand);
if (sc->ti_irq)
bus_release_resource(dev, SYS_RES_IRQ, 0, sc->ti_irq);
if (sc->ti_res) {
Remove ti_unit member variable in softc. While I'm here use PCIR_BAR macro.
2011-11-07 18:19:20 +00:00
bus_release_resource(dev, SYS_RES_MEMORY, PCIR_BAR(0),
Clean up locking and resource management for pci/if_* - Remove locking of the softc in the attach method, instead depending on bus_setup_intr being at the end of attach (delaying interrupt enable until after ether_ifattach is called) - Call *_detach directly in the error case of attach, depending on checking in detach to only free resources that were allocated. This puts all resource freeing in one place, avoiding thinkos that lead to memory leaks. - Add bus_child_present check to calls to *_stop in the detach method to be sure hw is present before touching its registers. - Remove bzero softc calls since device_t should do this for us. - dc: move interrupt allocation back where it was before. It was unnecessary to move it. This reverts part of 1.88 - rl: move irq allocation before ether_ifattach. Problems might have been caused by allocating the irq after enabling interrupts on the card. - rl: call rl_stop before ether_ifdetach - sf: call sf_stop before ether_ifdetach - sis: add missed free of sis_tag - sis: check errors from tag creation - sis: move dmamem_alloc and dmamap_load to happen at same time as tag creation - sk: remove duplicate initialization of sk_dev - ste: add missed bus_generic_detach - ti: call ti_stop before ether_ifdetach - ti: add missed error setting in ti_rdata alloc failure - vr: add missed error setting in I/O, memory mapping cases - xl: add missed error setting in I/O, memory mapping cases - xl: remove multi-level goto on attach failure - xl: move dmamem_alloc and dmamap_load to happen at same time as tag creation - Calls to free(9) are unconditional because it is valid to call free with a null pointer. Reviewed by: imp, mdodd
2003-03-31 17:29:43 +00:00
sc->ti_res);
}
In detach method, move if_free() after bus_teardown_intr().
2005-10-13 21:11:20 +00:00
if (ifp)
if_free(ifp);
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
if (sc->ti_membuf)
free(sc->ti_membuf, M_DEVBUF);
if (sc->ti_membuf2)
free(sc->ti_membuf2, M_DEVBUF);
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
mtx_destroy(&sc->ti_mtx);
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#ifdef TI_JUMBO_HDRSPLIT
/*
* If hdr_len is 0, that means that header splitting wasn't done on
* this packet for some reason. The two most likely reasons are that
* the protocol isn't a supported protocol for splitting, or this
* packet had a fragment offset that wasn't 0.
*
* The header length, if it is non-zero, will always be the length of
* the headers on the packet, but that length could be longer than the
* first mbuf. So we take the minimum of the two as the actual
Whitespace nits
2004-07-05 22:35:18 +00:00
* length.
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
*/
static __inline void
ti_hdr_split(struct mbuf *top, int hdr_len, int pkt_len, int idx)
{
int i = 0;
int lengths[4] = {0, 0, 0, 0};
struct mbuf *m, *mp;
if (hdr_len != 0)
top->m_len = min(hdr_len, top->m_len);
pkt_len -= top->m_len;
lengths[i++] = top->m_len;
mp = top;
for (m = top->m_next; m && pkt_len; m = m->m_next) {
m->m_len = m->m_ext.ext_size = min(m->m_len, pkt_len);
pkt_len -= m->m_len;
lengths[i++] = m->m_len;
mp = m;
}
#if 0
if (hdr_len != 0)
printf("got split packet: ");
else
printf("got non-split packet: ");
Whitespace nits
2004-07-05 22:35:18 +00:00
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
printf("%d,%d,%d,%d = %d\n", lengths[0],
lengths[1], lengths[2], lengths[3],
lengths[0] + lengths[1] + lengths[2] +
lengths[3]);
#endif
if (pkt_len)
panic("header splitting didn't");
Whitespace nits
2004-07-05 22:35:18 +00:00
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (m) {
m_freem(m);
mp->m_next = NULL;
}
if (mp->m_next != NULL)
panic("ti_hdr_split: last mbuf in chain should be null");
}
#endif /* TI_JUMBO_HDRSPLIT */
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
static void
ti_discard_std(struct ti_softc *sc, int i)
{
struct ti_rx_desc *r;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
r = &sc->ti_rdata.ti_rx_std_ring[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
r->ti_len = MCLBYTES - ETHER_ALIGN;
r->ti_type = TI_BDTYPE_RECV_BD;
r->ti_flags = 0;
r->ti_vlan_tag = 0;
r->ti_tcp_udp_cksum = 0;
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
r->ti_flags |= TI_BDFLAG_TCP_UDP_CKSUM | TI_BDFLAG_IP_CKSUM;
r->ti_idx = i;
}
static void
ti_discard_mini(struct ti_softc *sc, int i)
{
struct ti_rx_desc *r;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
r = &sc->ti_rdata.ti_rx_mini_ring[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
r->ti_len = MHLEN - ETHER_ALIGN;
r->ti_type = TI_BDTYPE_RECV_BD;
r->ti_flags = TI_BDFLAG_MINI_RING;
r->ti_vlan_tag = 0;
r->ti_tcp_udp_cksum = 0;
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
r->ti_flags |= TI_BDFLAG_TCP_UDP_CKSUM | TI_BDFLAG_IP_CKSUM;
r->ti_idx = i;
}
#ifndef TI_SF_BUF_JUMBO
static void
ti_discard_jumbo(struct ti_softc *sc, int i)
{
struct ti_rx_desc *r;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
r = &sc->ti_rdata.ti_rx_jumbo_ring[i];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
r->ti_len = MJUM9BYTES - ETHER_ALIGN;
r->ti_type = TI_BDTYPE_RECV_JUMBO_BD;
r->ti_flags = TI_BDFLAG_JUMBO_RING;
r->ti_vlan_tag = 0;
r->ti_tcp_udp_cksum = 0;
if (sc->ti_ifp->if_capenable & IFCAP_RXCSUM)
r->ti_flags |= TI_BDFLAG_TCP_UDP_CKSUM | TI_BDFLAG_IP_CKSUM;
r->ti_idx = i;
}
#endif
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Frame reception handling. This is called if there's a frame
* on the receive return list.
*
* Note: we have to be able to handle three possibilities here:
* 1) the frame is from the mini receive ring (can only happen)
* on Tigon 2 boards)
* 2) the frame is from the jumbo recieve ring
* 3) the frame is from the standard receive ring
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_rxeof(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ifnet *ifp;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
#ifdef TI_SF_BUF_JUMBO
style. No functional changes.
2011-11-04 18:28:10 +00:00
bus_dmamap_t map;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
#endif
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_cmd_desc cmd;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
int jumbocnt, minicnt, stdcnt, ti_len;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Drop the driver lock around calls to if_input to avoid a LOR when the packets are immediately returned for sending (e.g. when bridging or packet forwarding). There are more efficient ways to do this but for now use the least intrusive approach. Reviewed by: imp, rwatson
2003-11-14 19:00:32 +00:00
TI_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_std_ring_tag,
sc->ti_cdata.ti_rx_std_ring_map, BUS_DMASYNC_POSTWRITE);
if (ifp->if_mtu > ETHERMTU + ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)
bus_dmamap_sync(sc->ti_cdata.ti_rx_jumbo_ring_tag,
sc->ti_cdata.ti_rx_jumbo_ring_map, BUS_DMASYNC_POSTWRITE);
if (sc->ti_rdata.ti_rx_mini_ring != NULL)
bus_dmamap_sync(sc->ti_cdata.ti_rx_mini_ring_tag,
sc->ti_cdata.ti_rx_mini_ring_map, BUS_DMASYNC_POSTWRITE);
bus_dmamap_sync(sc->ti_cdata.ti_rx_return_ring_tag,
sc->ti_cdata.ti_rx_return_ring_map, BUS_DMASYNC_POSTREAD);
Track which ring was updated in RX handler and update only modified ring. This should reduce unnecessary register accesses.
2011-11-07 22:11:38 +00:00
jumbocnt = minicnt = stdcnt = 0;
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
while (sc->ti_rx_saved_considx != sc->ti_return_prodidx.ti_idx) {
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_rx_desc *cur_rx;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t rxidx;
Track which ring was updated in RX handler and update only modified ring. This should reduce unnecessary register accesses.
2011-11-07 22:11:38 +00:00
struct mbuf *m = NULL;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint16_t vlan_tag = 0;
style. No functional changes.
2011-11-04 18:28:10 +00:00
int have_tag = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
cur_rx =
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
&sc->ti_rdata.ti_rx_return_ring[sc->ti_rx_saved_considx];
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
rxidx = cur_rx->ti_idx;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_len = cur_rx->ti_len;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_INC(sc->ti_rx_saved_considx, TI_RETURN_RING_CNT);
if (cur_rx->ti_flags & TI_BDFLAG_VLAN_TAG) {
have_tag = 1;
Don't clear upper 4bits from VLAN tag information. It's responsibility of vlan(4) to extract VLAN id from the tag information and vlan(4) correctly handles it.
2011-11-04 21:42:13 +00:00
vlan_tag = cur_rx->ti_vlan_tag;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
if (cur_rx->ti_flags & TI_BDFLAG_JUMBO_RING) {
Track which ring was updated in RX handler and update only modified ring. This should reduce unnecessary register accesses.
2011-11-07 22:11:38 +00:00
jumbocnt++;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_INC(sc->ti_jumbo, TI_JUMBO_RX_RING_CNT);
m = sc->ti_cdata.ti_rx_jumbo_chain[rxidx];
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
#ifndef TI_SF_BUF_JUMBO
if (cur_rx->ti_flags & TI_BDFLAG_ERROR) {
ifp->if_ierrors++;
ti_discard_jumbo(sc, rxidx);
continue;
}
if (ti_newbuf_jumbo(sc, rxidx, NULL) != 0) {
ifp->if_iqdrops++;
ti_discard_jumbo(sc, rxidx);
continue;
}
m->m_len = ti_len;
#else /* !TI_SF_BUF_JUMBO */
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
sc->ti_cdata.ti_rx_jumbo_chain[rxidx] = NULL;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
map = sc->ti_cdata.ti_rx_jumbo_maps[rxidx];
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_jumbo_tag, map,
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
BUS_DMASYNC_POSTREAD);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_rx_jumbo_tag, map);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (cur_rx->ti_flags & TI_BDFLAG_ERROR) {
ifp->if_ierrors++;
ti_newbuf_jumbo(sc, sc->ti_jumbo, m);
continue;
}
Remove ti_refill_rx_rings() and associated stuff; replace dirty RX buffers in ti_rxeof() instead. This doesn't really seem to provide much in the way of a performance boost, and I'm pretty sure it can cause mbuf leakage in some extreme cases.
1999-07-05 20:19:41 +00:00
if (ti_newbuf_jumbo(sc, sc->ti_jumbo, NULL) == ENOBUFS) {
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ifp->if_iqdrops++;
Remove ti_refill_rx_rings() and associated stuff; replace dirty RX buffers in ti_rxeof() instead. This doesn't really seem to provide much in the way of a performance boost, and I'm pretty sure it can cause mbuf leakage in some extreme cases.
1999-07-05 20:19:41 +00:00
ti_newbuf_jumbo(sc, sc->ti_jumbo, m);
continue;
}
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#ifdef TI_JUMBO_HDRSPLIT
if (sc->ti_hdrsplit)
ti_hdr_split(m, TI_HOSTADDR(cur_rx->ti_addr),
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_len, rxidx);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
else
#endif /* TI_JUMBO_HDRSPLIT */
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
m_adj(m, ti_len - m->m_pkthdr.len);
#endif /* TI_SF_BUF_JUMBO */
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
} else if (cur_rx->ti_flags & TI_BDFLAG_MINI_RING) {
Track which ring was updated in RX handler and update only modified ring. This should reduce unnecessary register accesses.
2011-11-07 22:11:38 +00:00
minicnt++;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_INC(sc->ti_mini, TI_MINI_RX_RING_CNT);
m = sc->ti_cdata.ti_rx_mini_chain[rxidx];
if (cur_rx->ti_flags & TI_BDFLAG_ERROR) {
ifp->if_ierrors++;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_discard_mini(sc, rxidx);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
continue;
}
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
if (ti_newbuf_mini(sc, rxidx) != 0) {
ifp->if_iqdrops++;
ti_discard_mini(sc, rxidx);
Remove ti_refill_rx_rings() and associated stuff; replace dirty RX buffers in ti_rxeof() instead. This doesn't really seem to provide much in the way of a performance boost, and I'm pretty sure it can cause mbuf leakage in some extreme cases.
1999-07-05 20:19:41 +00:00
continue;
}
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
m->m_len = ti_len;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
} else {
Track which ring was updated in RX handler and update only modified ring. This should reduce unnecessary register accesses.
2011-11-07 22:11:38 +00:00
stdcnt++;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_INC(sc->ti_std, TI_STD_RX_RING_CNT);
m = sc->ti_cdata.ti_rx_std_chain[rxidx];
if (cur_rx->ti_flags & TI_BDFLAG_ERROR) {
ifp->if_ierrors++;
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
ti_discard_std(sc, rxidx);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
continue;
}
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
if (ti_newbuf_std(sc, rxidx) != 0) {
ifp->if_iqdrops++;
ti_discard_std(sc, rxidx);
Remove ti_refill_rx_rings() and associated stuff; replace dirty RX buffers in ti_rxeof() instead. This doesn't really seem to provide much in the way of a performance boost, and I'm pretty sure it can cause mbuf leakage in some extreme cases.
1999-07-05 20:19:41 +00:00
continue;
}
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
m->m_len = ti_len;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Retire 'options TI_PRIVATE_JUMBOS' and replace local jumbo allocator with UMA backed jumbo allocator by default. Previously ti(4) used sf_buf(9) interface for jumbo buffers but it was broken at this moment such that enabling jumbo frame caused instant panic. Due to the nature of sf_buf(9) it heavily relies on VM changes but it seems ti(4) was not received much blessing from VM gurus. I don't understand VM magic and implications used in driver either. Switching to UMA backed jumbo allocator like other network drivers will make jumbo frame work on ti(4). While I'm here, fully allocate all RX buffers. This means ti(4) now uses 512 RX buffer and 1024 mini RX buffers. To use sf_buf(9) interface for jumbo buffers, introduce a new 'options TI_SF_BUF_JUMBO'. If it is proven that sf_buf(9) is better for jumbo buffers, interesting developers can fix the issue in future. ti(4) still needs more bus_dma(9) cleanups and should use separate DMA tag/map for each ring(standard, jumbo, mini, command, event etc) but it should work on all platforms except PAE. Special thanks to Jay[1] who provided complete remote debugging access. Tested by: Jay Borkenhagen <jayb <> braeburn dot org > [1]
2011-11-08 18:23:02 +00:00
m->m_pkthdr.len = ti_len;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifp->if_ipackets++;
m->m_pkthdr.rcvif = ifp;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if (ifp->if_capenable & IFCAP_RXCSUM) {
if (cur_rx->ti_flags & TI_BDFLAG_IP_CKSUM) {
m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
if ((cur_rx->ti_ip_cksum ^ 0xffff) == 0)
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
}
if (cur_rx->ti_flags & TI_BDFLAG_TCP_UDP_CKSUM) {
m->m_pkthdr.csum_data =
cur_rx->ti_tcp_udp_cksum;
m->m_pkthdr.csum_flags |= CSUM_DATA_VALID;
}
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
o track either_ifattach/ether_ifdetach API changes o use if_input for input packet processing o don't strip the Ethernet header for input packets o use BPF_* macros bpf tapping o call ether_ioctl to handle default ioctl case o track vlan changes Reviewed by: many Approved by: re
2002-11-14 23:49:09 +00:00
* If we received a packet with a vlan tag,
* tag it before passing the packet upward.
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
*/
- Fix VLAN_INPUT_TAG() macro, so that it doesn't touch mtag in case if memory allocation failed. - Remove fourth argument from VLAN_INPUT_TAG(), that was used incorrectly in almost all drivers. Indicate failure with mbuf value of NULL. In collaboration with: yongari, ru, sam
2005-12-18 18:24:27 +00:00
if (have_tag) {
Move ethernet VLAN tags from mtags to its own mbuf packet header field m_pkthdr.ether_vlan. The presence of the M_VLANTAG flag on the mbuf signifies the presence and validity of its content. Drivers that support hardware VLAN tag stripping fill in the received VLAN tag (containing both vlan and priority information) into the ether_vtag mbuf packet header field: m->m_pkthdr.ether_vtag = vlan_id; /* ntohs()? */ m->m_flags |= M_VLANTAG; to mark the packet m with the specified VLAN tag. On output the driver should check the mbuf for the M_VLANTAG flag to see if a VLAN tag is present and valid: if (m->m_flags & M_VLANTAG) { ... = m->m_pkthdr.ether_vtag; /* htons()? */ ... pass tag to hardware ... } VLAN tags are stored in host byte order. Byte swapping may be necessary. (Note: This driver conversion was mechanic and did not add or remove any byte swapping in the drivers.) Remove zone_mtag_vlan UMA zone and MTAG_VLAN definition. No more tag memory allocation have to be done. Reviewed by: thompsa, yar Sponsored by: TCP/IP Optimization Fundraise 2005
2006-09-17 13:33:30 +00:00
m->m_pkthdr.ether_vtag = vlan_tag;
m->m_flags |= M_VLANTAG;
- Fix VLAN_INPUT_TAG() macro, so that it doesn't touch mtag in case if memory allocation failed. - Remove fourth argument from VLAN_INPUT_TAG(), that was used incorrectly in almost all drivers. Indicate failure with mbuf value of NULL. In collaboration with: yongari, ru, sam
2005-12-18 18:24:27 +00:00
}
Drop the driver lock around calls to if_input to avoid a LOR when the packets are immediately returned for sending (e.g. when bridging or packet forwarding). There are more efficient ways to do this but for now use the least intrusive approach. Reviewed by: imp, rwatson
2003-11-14 19:00:32 +00:00
TI_UNLOCK(sc);
o track either_ifattach/ether_ifdetach API changes o use if_input for input packet processing o don't strip the Ethernet header for input packets o use BPF_* macros bpf tapping o call ether_ioctl to handle default ioctl case o track vlan changes Reviewed by: many Approved by: re
2002-11-14 23:49:09 +00:00
(*ifp->if_input)(ifp, m);
Drop the driver lock around calls to if_input to avoid a LOR when the packets are immediately returned for sending (e.g. when bridging or packet forwarding). There are more efficient ways to do this but for now use the least intrusive approach. Reviewed by: imp, rwatson
2003-11-14 19:00:32 +00:00
TI_LOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_rx_return_ring_tag,
sc->ti_cdata.ti_rx_return_ring_map, BUS_DMASYNC_PREREAD);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Only necessary on the Tigon 1. */
if (sc->ti_hwrev == TI_HWREV_TIGON)
CSR_WRITE_4(sc, TI_GCR_RXRETURNCONS_IDX,
sc->ti_rx_saved_considx);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (stdcnt > 0) {
bus_dmamap_sync(sc->ti_cdata.ti_rx_std_ring_tag,
sc->ti_cdata.ti_rx_std_ring_map, BUS_DMASYNC_PREWRITE);
Track which ring was updated in RX handler and update only modified ring. This should reduce unnecessary register accesses.
2011-11-07 22:11:38 +00:00
TI_UPDATE_STDPROD(sc, sc->ti_std);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
}
if (minicnt > 0) {
bus_dmamap_sync(sc->ti_cdata.ti_rx_mini_ring_tag,
sc->ti_cdata.ti_rx_mini_ring_map, BUS_DMASYNC_PREWRITE);
Track which ring was updated in RX handler and update only modified ring. This should reduce unnecessary register accesses.
2011-11-07 22:11:38 +00:00
TI_UPDATE_MINIPROD(sc, sc->ti_mini);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
}
if (jumbocnt > 0) {
bus_dmamap_sync(sc->ti_cdata.ti_rx_jumbo_ring_tag,
sc->ti_cdata.ti_rx_jumbo_ring_map, BUS_DMASYNC_PREWRITE);
Track which ring was updated in RX handler and update only modified ring. This should reduce unnecessary register accesses.
2011-11-07 22:11:38 +00:00
TI_UPDATE_JUMBOPROD(sc, sc->ti_jumbo);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_txeof(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_txdesc *txd;
struct ti_tx_desc txdesc;
struct ti_tx_desc *cur_tx = NULL;
struct ifnet *ifp;
int idx;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
txd = STAILQ_FIRST(&sc->ti_cdata.ti_txbusyq);
if (txd == NULL)
return;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_rdata.ti_tx_ring != NULL)
bus_dmamap_sync(sc->ti_cdata.ti_tx_ring_tag,
sc->ti_cdata.ti_tx_ring_map, BUS_DMASYNC_POSTWRITE);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Go through our tx ring and free mbufs for those
* frames that have been sent.
*/
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
for (idx = sc->ti_tx_saved_considx; idx != sc->ti_tx_considx.ti_idx;
TI_INC(idx, TI_TX_RING_CNT)) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON) {
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
ti_mem_read(sc, TI_TX_RING_BASE + idx * sizeof(txdesc),
sizeof(txdesc), &txdesc);
cur_tx = &txdesc;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
} else
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
cur_tx = &sc->ti_rdata.ti_tx_ring[idx];
Add a transmit descriptor usage counter and use it to absolutely, positively not let ti_encap() fill up the TX ring all the way and wrap around. This fixes a potential transmit lockup where a really fast machine (or particular TX traffic pattern) can overrun the end of the ring. Reported by: John Plevyak <jplevyak@inktomi.com>
1999-06-19 00:36:56 +00:00
sc->ti_txcnt--;
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
if ((cur_tx->ti_flags & TI_BDFLAG_END) == 0)
continue;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_tx_tag, txd->tx_dmamap,
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
BUS_DMASYNC_POSTWRITE);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_tx_tag, txd->tx_dmamap);
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
ifp->if_opackets++;
m_freem(txd->tx_m);
txd->tx_m = NULL;
STAILQ_REMOVE_HEAD(&sc->ti_cdata.ti_txbusyq, tx_q);
STAILQ_INSERT_TAIL(&sc->ti_cdata.ti_txfreeq, txd, tx_q);
txd = STAILQ_FIRST(&sc->ti_cdata.ti_txbusyq);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
sc->ti_tx_saved_considx = idx;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_txcnt == 0)
sc->ti_timer = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_intr(void *xsc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
struct ifnet *ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
sc = xsc;
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
TI_LOCK(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Make sure this is really our interrupt. */
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
if (!(CSR_READ_4(sc, TI_MISC_HOST_CTL) & TI_MHC_INTSTATE)) {
TI_UNLOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
return;
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Ack interrupt and stop others from occuring. */
CSR_WRITE_4(sc, TI_MB_HOSTINTR, 1);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_status_tag,
sc->ti_cdata.ti_status_map, BUS_DMASYNC_POSTREAD);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Check RX return ring producer/consumer */
ti_rxeof(sc);
/* Check TX ring producer/consumer */
ti_txeof(sc);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_status_tag,
sc->ti_cdata.ti_status_map, BUS_DMASYNC_PREREAD);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
ti_handle_events(sc);
Implement altq(4) support. While I'm here fix a logic error in r227098 where it didn't re-enable interrupts when TX queue is empty.
2011-11-04 23:34:54 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Because ti(4) drops a driver lock in RX handler, check whether driver is still running before re-enabling interrupts.
2011-11-04 23:09:57 +00:00
/* Re-enable interrupts. */
CSR_WRITE_4(sc, TI_MB_HOSTINTR, 0);
Implement altq(4) support. While I'm here fix a logic error in r227098 where it didn't re-enable interrupts when TX queue is empty.
2011-11-04 23:34:54 +00:00
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
ti_start_locked(ifp);
Because ti(4) drops a driver lock in RX handler, check whether driver is still running before re-enabling interrupts.
2011-11-04 23:09:57 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
TI_UNLOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_stats_update(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ifnet *ifp;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
struct ti_stats *s;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_stat_ticks == 0)
return;
bus_dmamap_sync(sc->ti_cdata.ti_gib_tag, sc->ti_cdata.ti_gib_map,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
BUS_DMASYNC_POSTREAD);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
s = &sc->ti_rdata.ti_info->ti_stats;
ifp->if_collisions += (s->dot3StatsSingleCollisionFrames +
s->dot3StatsMultipleCollisionFrames +
s->dot3StatsExcessiveCollisions + s->dot3StatsLateCollisions) -
ifp->if_collisions;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_gib_tag, sc->ti_cdata.ti_gib_map,
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
BUS_DMASYNC_PREREAD);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
/*
* Encapsulate an mbuf chain in the tx ring by coupling the mbuf data
* pointers to descriptors.
*/
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_encap(struct ti_softc *sc, struct mbuf **m_head)
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_txdesc *txd;
struct ti_tx_desc *f;
struct ti_tx_desc txdesc;
struct mbuf *m;
bus_dma_segment_t txsegs[TI_MAXTXSEGS];
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint16_t csum_flags;
style. No functional changes.
2011-11-04 18:28:10 +00:00
int error, frag, i, nseg;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
if ((txd = STAILQ_FIRST(&sc->ti_cdata.ti_txfreeq)) == NULL)
return (ENOBUFS);
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_load_mbuf_sg(sc->ti_cdata.ti_tx_tag, txd->tx_dmamap,
Fix invalid reference of mbuf chains. Use proper pointer dereference to inform modified mbuf chains to caller. While I'm here perform checksum offload setup after loading DMA maps. In collaboration with: glebius
2006-08-12 01:30:38 +00:00
*m_head, txsegs, &nseg, 0);
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
if (error == EFBIG) {
Mechanically substitute flags from historic mbuf allocator with malloc(9) flags in sys/dev.
2012-12-04 09:32:43 +00:00
m = m_defrag(*m_head, M_NOWAIT);
Fix invalid reference of mbuf chains. Use proper pointer dereference to inform modified mbuf chains to caller. While I'm here perform checksum offload setup after loading DMA maps. In collaboration with: glebius
2006-08-12 01:30:38 +00:00
if (m == NULL) {
m_freem(*m_head);
*m_head = NULL;
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
return (ENOMEM);
}
Fix invalid reference of mbuf chains. Use proper pointer dereference to inform modified mbuf chains to caller. While I'm here perform checksum offload setup after loading DMA maps. In collaboration with: glebius
2006-08-12 01:30:38 +00:00
*m_head = m;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
error = bus_dmamap_load_mbuf_sg(sc->ti_cdata.ti_tx_tag,
Fix invalid reference of mbuf chains. Use proper pointer dereference to inform modified mbuf chains to caller. While I'm here perform checksum offload setup after loading DMA maps. In collaboration with: glebius
2006-08-12 01:30:38 +00:00
txd->tx_dmamap, *m_head, txsegs, &nseg, 0);
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
if (error) {
Fix invalid reference of mbuf chains. Use proper pointer dereference to inform modified mbuf chains to caller. While I'm here perform checksum offload setup after loading DMA maps. In collaboration with: glebius
2006-08-12 01:30:38 +00:00
m_freem(*m_head);
*m_head = NULL;
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
return (error);
}
} else if (error != 0)
return (error);
if (nseg == 0) {
Fix invalid reference of mbuf chains. Use proper pointer dereference to inform modified mbuf chains to caller. While I'm here perform checksum offload setup after loading DMA maps. In collaboration with: glebius
2006-08-12 01:30:38 +00:00
m_freem(*m_head);
*m_head = NULL;
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
return (EIO);
Fix a serious regression from the busdma conversion. Check to make sure that we don't overrun the tx descriptor ring before actually trying to overrun it.
2005-12-28 08:14:35 +00:00
}
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
if (sc->ti_txcnt + nseg >= TI_TX_RING_CNT) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_unload(sc->ti_cdata.ti_tx_tag, txd->tx_dmamap);
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
return (ENOBUFS);
}
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_tx_tag, txd->tx_dmamap,
BUS_DMASYNC_PREWRITE);
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
Fix invalid reference of mbuf chains. Use proper pointer dereference to inform modified mbuf chains to caller. While I'm here perform checksum offload setup after loading DMA maps. In collaboration with: glebius
2006-08-12 01:30:38 +00:00
m = *m_head;
csum_flags = 0;
Remove TCP/UDP checksum offloading feature for IP fragmented datagrams. Traditionally upper stack fragmented packets without computing TCP/UDP checksum and these datagrams were passed to driver. But there are chances that other packets slip into the interface queue in SMP world. If this happens firmware running on MIPS 4000 processor in the controller would see mixed packets and it shall send out corrupted packets. While I'm here simplify checksum offloading setup. MFC After: 1 week
2012-11-01 05:39:21 +00:00
if (m->m_pkthdr.csum_flags & CSUM_IP)
csum_flags |= TI_BDFLAG_IP_CKSUM;
if (m->m_pkthdr.csum_flags & (CSUM_TCP | CSUM_UDP))
csum_flags |= TI_BDFLAG_TCP_UDP_CKSUM;
Fix invalid reference of mbuf chains. Use proper pointer dereference to inform modified mbuf chains to caller. While I'm here perform checksum offload setup after loading DMA maps. In collaboration with: glebius
2006-08-12 01:30:38 +00:00
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
frag = sc->ti_tx_saved_prodidx;
for (i = 0; i < nseg; i++) {
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON) {
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
bzero(&txdesc, sizeof(txdesc));
f = &txdesc;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
} else
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
f = &sc->ti_rdata.ti_tx_ring[frag];
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
ti_hostaddr64(&f->ti_addr, txsegs[i].ds_addr);
f->ti_len = txsegs[i].ds_len;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
f->ti_flags = csum_flags;
Move ethernet VLAN tags from mtags to its own mbuf packet header field m_pkthdr.ether_vlan. The presence of the M_VLANTAG flag on the mbuf signifies the presence and validity of its content. Drivers that support hardware VLAN tag stripping fill in the received VLAN tag (containing both vlan and priority information) into the ether_vtag mbuf packet header field: m->m_pkthdr.ether_vtag = vlan_id; /* ntohs()? */ m->m_flags |= M_VLANTAG; to mark the packet m with the specified VLAN tag. On output the driver should check the mbuf for the M_VLANTAG flag to see if a VLAN tag is present and valid: if (m->m_flags & M_VLANTAG) { ... = m->m_pkthdr.ether_vtag; /* htons()? */ ... pass tag to hardware ... } VLAN tags are stored in host byte order. Byte swapping may be necessary. (Note: This driver conversion was mechanic and did not add or remove any byte swapping in the drivers.) Remove zone_mtag_vlan UMA zone and MTAG_VLAN definition. No more tag memory allocation have to be done. Reviewed by: thompsa, yar Sponsored by: TCP/IP Optimization Fundraise 2005
2006-09-17 13:33:30 +00:00
if (m->m_flags & M_VLANTAG) {
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
f->ti_flags |= TI_BDFLAG_VLAN_TAG;
Don't clear upper 4bits from VLAN tag information. It's responsibility of vlan(4) to extract VLAN id from the tag information and vlan(4) correctly handles it.
2011-11-04 21:42:13 +00:00
f->ti_vlan_tag = m->m_pkthdr.ether_vtag;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
} else {
f->ti_vlan_tag = 0;
}
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON)
ti_mem_write(sc, TI_TX_RING_BASE + frag *
sizeof(txdesc), sizeof(txdesc), &txdesc);
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
TI_INC(frag, TI_TX_RING_CNT);
}
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
sc->ti_tx_saved_prodidx = frag;
/* set TI_BDFLAG_END on the last descriptor */
frag = (frag + TI_TX_RING_CNT - 1) % TI_TX_RING_CNT;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON) {
txdesc.ti_flags |= TI_BDFLAG_END;
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
ti_mem_write(sc, TI_TX_RING_BASE + frag * sizeof(txdesc),
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
sizeof(txdesc), &txdesc);
} else
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
sc->ti_rdata.ti_tx_ring[frag].ti_flags |= TI_BDFLAG_END;
Fix the Tigon I/II driver to support 64-bit DMA. In the process, convert it to use busdma. Unlike most of the other drivers, but similar to the if_em driver, pre-allocate the dmamaps at init time instead of allocating them on the fly when descriptors need to be filled. This isn't ideal right now because a map is allocated for every descriptor slot in the tx, rx, mini, and jumbo rings (which is a lot!) in order to simplify the bookkeeping, even though the driver might support filling only a subset of those slots. Luckily, maps are typically NULL on i386 and amd64, so the cost isn't very high. It could be an issue with sparc64, but the driver isn't endian clean either, and that is a much bigger problem to solve first. Note that jumbo frame support is under-tested, and I'm not even sure if it till really works correctly given the evil VM magic that is does. The changes here attempt to preserve the existing semanitcs. Thanks to Martin Nillson for contributing the Netgear card for this work. MFC-After: 3 weeks
2005-12-14 00:03:41 +00:00
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
STAILQ_REMOVE_HEAD(&sc->ti_cdata.ti_txfreeq, tx_q);
STAILQ_INSERT_TAIL(&sc->ti_cdata.ti_txbusyq, txd, tx_q);
txd->tx_m = m;
sc->ti_txcnt += nseg;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_start(struct ifnet *ifp)
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
sc = ifp->if_softc;
TI_LOCK(sc);
ti_start_locked(ifp);
TI_UNLOCK(sc);
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Main transmit routine. To avoid having to do mbuf copies, we put pointers
* to the mbuf data regions directly in the transmit descriptors.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_start_locked(struct ifnet *ifp)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
struct mbuf *m_head = NULL;
int enq = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
sc = ifp->if_softc;
Implement altq(4) support. While I'm here fix a logic error in r227098 where it didn't re-enable interrupts when TX queue is empty.
2011-11-04 23:34:54 +00:00
for (; !IFQ_DRV_IS_EMPTY(&ifp->if_snd) &&
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
sc->ti_txcnt < (TI_TX_RING_CNT - 16);) {
Implement altq(4) support. While I'm here fix a logic error in r227098 where it didn't re-enable interrupts when TX queue is empty.
2011-11-04 23:34:54 +00:00
IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (m_head == NULL)
break;
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
/*
* XXX
* safety overkill. If this is a fragmented packet chain
* with delayed TCP/UDP checksums, then only encapsulate
* it if we have enough descriptors to handle the entire
* chain at once.
* (paranoia -- may not actually be needed)
*/
if (m_head->m_flags & M_FIRSTFRAG &&
m_head->m_pkthdr.csum_flags & (CSUM_DELAY_DATA)) {
if ((TI_TX_RING_CNT - sc->ti_txcnt) <
m_head->m_pkthdr.csum_data + 16) {
Implement altq(4) support. While I'm here fix a logic error in r227098 where it didn't re-enable interrupts when TX queue is empty.
2011-11-04 23:34:54 +00:00
IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Add support for offloading IP/TCP/UDP checksums to NIC hardware which supports them.
2000-03-27 19:14:27 +00:00
break;
}
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Pack the data into the transmit ring. If we
* don't have room, set the OACTIVE flag and wait
* for the NIC to drain the ring.
*/
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
if (ti_encap(sc, &m_head)) {
if (m_head == NULL)
break;
Implement altq(4) support. While I'm here fix a logic error in r227098 where it didn't re-enable interrupts when TX queue is empty.
2011-11-04 23:34:54 +00:00
IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
}
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
enq++;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* If there's a BPF listener, bounce a copy of this frame
* to him.
*/
Catch up the rest of the drivers with the ether_vlan_mtap modifications. If these drivers are setting M_VLANTAG because they are stripping the layer 2 802.1Q headers, then they need to be re-inserting them so any bpf(4) peers can properly decode them. It should be noted that this is compiled tested only. MFC after: 3 weeks
2007-03-04 03:38:08 +00:00
ETHER_BPF_MTAP(ifp, m_head);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
if (enq > 0) {
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (sc->ti_rdata.ti_tx_ring != NULL)
bus_dmamap_sync(sc->ti_cdata.ti_tx_ring_tag,
sc->ti_cdata.ti_tx_ring_map, BUS_DMASYNC_PREWRITE);
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
/* Transmit */
CSR_WRITE_4(sc, TI_MB_SENDPROD_IDX, sc->ti_tx_saved_prodidx);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
/*
* Set a timeout in case the chip goes out to lunch.
*/
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
sc->ti_timer = 5;
- Tx side bus_dmamap_load_mbuf_sg(9) support. This reduces bookkeeping requiried to keep consistent softc state before/after callback function invocation and supposed to be sligntly faster than previous one as it wouldn't incur callback overhead. With this change callback function was gone. - Decrease TI_MAXTXSEGS to 32 from 128. It seems that most mbuf chain length is less than 32 and it would be re-packed with m_defrag(9) if its chain length is larger than TI_MAXTXSEGS. This would protect ti(4) against possible kernel stack overflow when txsegs[] is put on stack. Alternatively, we can embed the txsegs[] into softc. However, that would waste memory and make Tx/Rx speration hard when we want to sperate Tx/Rx handlers to optimize locking. - Fix dma map tracking used in Tx path. Previously it used the dma map of the last mbuf chain in ti_txeof() which was incorrect as ti(4) used dma map of the first mbuf chain when it loads a mbuf chain with bus_dmamap_load_mbuf(9). Correct the bug by introducing queues that keep track of active/inactive dma maps/mbuf chain. - Use ti_txcnt to check whether driver need to set watchdog timer instead of blidnly clearing the timer in ti_txeof(). - Remove the 3rd arg. of ti_encap(). Since ti(4) now caches the last descriptor index(ti_tx_saved_prodidx) used in Tx there is no need to pass it as a fuction arg. - Change data type of producer/consumer index to int from u_int16_t in order to remove implicit type conversions in Tx/Rx handlers. - Check interface queue before getting a mbuf chain to reduce locking overhead. - Check number of available Tx descriptores to be 16 or higher in ti_start(). This wouldn't protect Tx descriptor shortage but it would reduce number of bus_dmamap_unload(9) calls in ti_encap() when we are about to running out of Tx descriptors. - Command NIC to send packets ony when the driver really has packets enqueued. Previously it always set TI_MB_SENDPROD_IDX which would command NIC to DMA Tx descriptors into NIC local memory regardless of Tx descriptor changes. Reviewed by: scottl
2006-01-03 06:14:07 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_init(void *xsc)
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
sc = xsc;
TI_LOCK(sc);
ti_init_locked(sc);
TI_UNLOCK(sc);
}
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_init_locked(void *xsc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc = xsc;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
o Remove unnecessary controller reinitialization. o Do not blindly UP controller when MTU is changed. Reinitialize controller only if driver is running. o Remove useless ti_stop() in ti_watchdog() since ti_init_locked() always invokes ti_stop().
2011-11-07 18:26:38 +00:00
if (sc->ti_ifp->if_drv_flags & IFF_DRV_RUNNING)
return;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Cancel pending I/O and flush buffers. */
ti_stop(sc);
/* Init the gen info block, ring control blocks and firmware. */
if (ti_gibinit(sc)) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->ti_dev, "initialization failure\n");
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
return;
}
}
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
static void ti_init2(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_cmd_desc cmd;
struct ifnet *ifp;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint8_t *ea;
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ifmedia *ifm;
int tmp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Specify MTU and interface index. */
Remove ti_unit member variable in softc. While I'm here use PCIR_BAR macro.
2011-11-07 18:19:20 +00:00
CSR_WRITE_4(sc, TI_GCR_IFINDEX, device_get_unit(sc->ti_dev));
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_IFMTU, ifp->if_mtu +
Enable IFCAP_VLAN_MTU and increase MTU for it. Reviewed by: wpaul
2003-08-05 02:34:35 +00:00
ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_DO_CMD(TI_CMD_UPDATE_GENCOM, 0, 0);
/* Load our MAC address. */
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
ea = IF_LLADDR(sc->ti_ifp);
CSR_WRITE_4(sc, TI_GCR_PAR0, (ea[0] << 8) | ea[1]);
CSR_WRITE_4(sc, TI_GCR_PAR1,
(ea[2] << 24) | (ea[3] << 16) | (ea[4] << 8) | ea[5]);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_DO_CMD(TI_CMD_SET_MAC_ADDR, 0, 0);
/* Enable or disable promiscuous mode as needed. */
if (ifp->if_flags & IFF_PROMISC) {
TI_DO_CMD(TI_CMD_SET_PROMISC_MODE, TI_CMD_CODE_PROMISC_ENB, 0);
} else {
TI_DO_CMD(TI_CMD_SET_PROMISC_MODE, TI_CMD_CODE_PROMISC_DIS, 0);
}
/* Program multicast filter. */
ti_setmulti(sc);
/*
* If this is a Tigon 1, we should tell the
* firmware to use software packet filtering.
*/
if (sc->ti_hwrev == TI_HWREV_TIGON) {
TI_DO_CMD(TI_CMD_FDR_FILTERING, TI_CMD_CODE_FILT_ENB, 0);
}
/* Init RX ring. */
Show RX buffer allocation failure and do not blindly send alive message to firmware. Probably the correct way for this error is to send a TI_CMD_CODE_STACK_DOWN message to firmware and let firmware handle the rest.
2011-11-07 22:47:25 +00:00
if (ti_init_rx_ring_std(sc) != 0) {
/* XXX */
device_printf(sc->ti_dev, "no memory for std Rx buffers.\n");
return;
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Init jumbo RX ring. */
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
if (ifp->if_mtu > ETHERMTU + ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN) {
Show RX buffer allocation failure and do not blindly send alive message to firmware. Probably the correct way for this error is to send a TI_CMD_CODE_STACK_DOWN message to firmware and let firmware handle the rest.
2011-11-07 22:47:25 +00:00
if (ti_init_rx_ring_jumbo(sc) != 0) {
/* XXX */
device_printf(sc->ti_dev,
"no memory for jumbo Rx buffers.\n");
return;
}
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* If this is a Tigon 2, we can also configure the
* mini ring.
*/
Show RX buffer allocation failure and do not blindly send alive message to firmware. Probably the correct way for this error is to send a TI_CMD_CODE_STACK_DOWN message to firmware and let firmware handle the rest.
2011-11-07 22:47:25 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON_II) {
if (ti_init_rx_ring_mini(sc) != 0) {
/* XXX */
device_printf(sc->ti_dev,
"no memory for mini Rx buffers.\n");
return;
}
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_RXRETURNCONS_IDX, 0);
sc->ti_rx_saved_considx = 0;
/* Init TX ring. */
ti_init_tx_ring(sc);
/* Tell firmware we're alive. */
TI_DO_CMD(TI_CMD_HOST_STATE, TI_CMD_CODE_STACK_UP, 0);
/* Enable host interrupts. */
CSR_WRITE_4(sc, TI_MB_HOSTINTR, 0);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
callout_reset(&sc->ti_watchdog, hz, ti_watchdog, sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/*
* Make sure to set media properly. We have to do this
* here since we have to issue commands in order to set
* the link negotiation and we can't issue commands until
* the firmware is running.
*/
ifm = &sc->ifmedia;
tmp = ifm->ifm_media;
ifm->ifm_media = ifm->ifm_cur->ifm_media;
Introduce ti_ifmedia_upd_locked() to use in driver initialization and add missing driver lock for both ti_ifmedia_upd() and ti_ifmedia_sts().
2011-11-04 21:30:46 +00:00
ti_ifmedia_upd_locked(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifm->ifm_media = tmp;
}
/*
* Set media options.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_ifmedia_upd(struct ifnet *ifp)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
Introduce ti_ifmedia_upd_locked() to use in driver initialization and add missing driver lock for both ti_ifmedia_upd() and ti_ifmedia_sts().
2011-11-04 21:30:46 +00:00
int error;
sc = ifp->if_softc;
TI_LOCK(sc);
error = ti_ifmedia_upd(ifp);
TI_UNLOCK(sc);
return (error);
}
static int
ti_ifmedia_upd_locked(struct ti_softc *sc)
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ifmedia *ifm;
struct ti_cmd_desc cmd;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t flowctl;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifm = &sc->ifmedia;
if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (EINVAL);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
flowctl = 0;
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
switch (IFM_SUBTYPE(ifm->ifm_media)) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
case IFM_AUTO:
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* Transmit flow control doesn't work on the Tigon 1.
*/
flowctl = TI_GLNK_RX_FLOWCTL_Y;
/*
* Transmit flow control can also cause problems on the
* Tigon 2, apparantly with both the copper and fiber
* boards. The symptom is that the interface will just
* hang. This was reproduced with Alteon 180 switches.
*/
#if 0
if (sc->ti_hwrev != TI_HWREV_TIGON)
Whitespace nits
2004-07-05 22:35:18 +00:00
flowctl |= TI_GLNK_TX_FLOWCTL_Y;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#endif
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_GLINK, TI_GLNK_PREF|TI_GLNK_1000MB|
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
TI_GLNK_FULL_DUPLEX| flowctl |
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_GLNK_AUTONEGENB|TI_GLNK_ENB);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
flowctl = TI_LNK_RX_FLOWCTL_Y;
#if 0
if (sc->ti_hwrev != TI_HWREV_TIGON)
flowctl |= TI_LNK_TX_FLOWCTL_Y;
#endif
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_LINK, TI_LNK_100MB|TI_LNK_10MB|
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
TI_LNK_FULL_DUPLEX|TI_LNK_HALF_DUPLEX| flowctl |
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_LNK_AUTONEGENB|TI_LNK_ENB);
TI_DO_CMD(TI_CMD_LINK_NEGOTIATION,
TI_CMD_CODE_NEGOTIATE_BOTH, 0);
break;
case IFM_1000_SX:
Follow NetBSD and s/IFM_1000_TX/IFM_1000_T/
2002-04-28 20:34:20 +00:00
case IFM_1000_T:
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
flowctl = TI_GLNK_RX_FLOWCTL_Y;
#if 0
if (sc->ti_hwrev != TI_HWREV_TIGON)
Whitespace nits
2004-07-05 22:35:18 +00:00
flowctl |= TI_GLNK_TX_FLOWCTL_Y;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
#endif
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_GLINK, TI_GLNK_PREF|TI_GLNK_1000MB|
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
flowctl |TI_GLNK_ENB);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_LINK, 0);
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) {
TI_SETBIT(sc, TI_GCR_GLINK, TI_GLNK_FULL_DUPLEX);
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_DO_CMD(TI_CMD_LINK_NEGOTIATION,
TI_CMD_CODE_NEGOTIATE_GIGABIT, 0);
break;
case IFM_100_FX:
case IFM_10_FL:
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
case IFM_100_TX:
case IFM_10_T:
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
flowctl = TI_LNK_RX_FLOWCTL_Y;
#if 0
if (sc->ti_hwrev != TI_HWREV_TIGON)
flowctl |= TI_LNK_TX_FLOWCTL_Y;
#endif
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
CSR_WRITE_4(sc, TI_GCR_GLINK, 0);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
CSR_WRITE_4(sc, TI_GCR_LINK, TI_LNK_ENB|TI_LNK_PREF|flowctl);
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
if (IFM_SUBTYPE(ifm->ifm_media) == IFM_100_FX ||
IFM_SUBTYPE(ifm->ifm_media) == IFM_100_TX) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
TI_SETBIT(sc, TI_GCR_LINK, TI_LNK_100MB);
} else {
TI_SETBIT(sc, TI_GCR_LINK, TI_LNK_10MB);
}
if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) {
TI_SETBIT(sc, TI_GCR_LINK, TI_LNK_FULL_DUPLEX);
} else {
TI_SETBIT(sc, TI_GCR_LINK, TI_LNK_HALF_DUPLEX);
}
TI_DO_CMD(TI_CMD_LINK_NEGOTIATION,
TI_CMD_CODE_NEGOTIATE_10_100, 0);
break;
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Report current media status.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t media = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
sc = ifp->if_softc;
Introduce ti_ifmedia_upd_locked() to use in driver initialization and add missing driver lock for both ti_ifmedia_upd() and ti_ifmedia_sts().
2011-11-04 21:30:46 +00:00
TI_LOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifmr->ifm_status = IFM_AVALID;
ifmr->ifm_active = IFM_ETHER;
Introduce ti_ifmedia_upd_locked() to use in driver initialization and add missing driver lock for both ti_ifmedia_upd() and ti_ifmedia_sts().
2011-11-04 21:30:46 +00:00
if (sc->ti_linkstat == TI_EV_CODE_LINK_DOWN) {
TI_UNLOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
return;
Introduce ti_ifmedia_upd_locked() to use in driver initialization and add missing driver lock for both ti_ifmedia_upd() and ti_ifmedia_sts().
2011-11-04 21:30:46 +00:00
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifmr->ifm_status |= IFM_ACTIVE;
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
if (sc->ti_linkstat == TI_EV_CODE_GIG_LINK_UP) {
media = CSR_READ_4(sc, TI_GCR_GLINK_STAT);
if (sc->ti_copper)
Follow NetBSD and s/IFM_1000_TX/IFM_1000_T/
2002-04-28 20:34:20 +00:00
ifmr->ifm_active |= IFM_1000_T;
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
else
ifmr->ifm_active |= IFM_1000_SX;
if (media & TI_GLNK_FULL_DUPLEX)
ifmr->ifm_active |= IFM_FDX;
else
ifmr->ifm_active |= IFM_HDX;
} else if (sc->ti_linkstat == TI_EV_CODE_LINK_UP) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
media = CSR_READ_4(sc, TI_GCR_LINK_STAT);
Update the Tigon driver to support 1000baseTX gigE over copper AceNIC cards. This basically involves switching to the 12.4.13 firmware, plus a couple of minor tweaks to the driver. Also changed the jumbo buffer allocation scheme just a little to avoid 'failed to allocate jumbo buffer' conditions in certain cases.
2000-07-20 22:24:43 +00:00
if (sc->ti_copper) {
if (media & TI_LNK_100MB)
ifmr->ifm_active |= IFM_100_TX;
if (media & TI_LNK_10MB)
ifmr->ifm_active |= IFM_10_T;
} else {
if (media & TI_LNK_100MB)
ifmr->ifm_active |= IFM_100_FX;
if (media & TI_LNK_10MB)
ifmr->ifm_active |= IFM_10_FL;
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (media & TI_LNK_FULL_DUPLEX)
ifmr->ifm_active |= IFM_FDX;
if (media & TI_LNK_HALF_DUPLEX)
ifmr->ifm_active |= IFM_HDX;
}
Introduce ti_ifmedia_upd_locked() to use in driver initialization and add missing driver lock for both ti_ifmedia_upd() and ti_ifmedia_sts().
2011-11-04 21:30:46 +00:00
TI_UNLOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc = ifp->if_softc;
struct ifreq *ifr = (struct ifreq *) data;
struct ti_cmd_desc cmd;
int mask, error = 0;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
switch (command) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
case SIOCSIFMTU:
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK(sc);
Do not allow changing MTU to be less than the minimum.
2011-11-07 22:58:49 +00:00
if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > TI_JUMBO_MTU)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
error = EINVAL;
else {
ifp->if_mtu = ifr->ifr_mtu;
o Remove unnecessary controller reinitialization. o Do not blindly UP controller when MTU is changed. Reinitialize controller only if driver is running. o Remove useless ti_stop() in ti_watchdog() since ti_init_locked() always invokes ti_stop().
2011-11-07 18:26:38 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
ti_init_locked(sc);
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_UNLOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
case SIOCSIFFLAGS:
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
if (ifp->if_flags & IFF_UP) {
/*
* If only the state of the PROMISC flag changed,
* then just use the 'set promisc mode' command
* instead of reinitializing the entire NIC. Doing
* a full re-init means reloading the firmware and
* waiting for it to start up, which may take a
* second or two.
*/
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifp->if_flags & IFF_PROMISC &&
!(sc->ti_if_flags & IFF_PROMISC)) {
TI_DO_CMD(TI_CMD_SET_PROMISC_MODE,
TI_CMD_CODE_PROMISC_ENB, 0);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
} else if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
!(ifp->if_flags & IFF_PROMISC) &&
sc->ti_if_flags & IFF_PROMISC) {
TI_DO_CMD(TI_CMD_SET_PROMISC_MODE,
TI_CMD_CODE_PROMISC_DIS, 0);
} else
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
ti_init_locked(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
} else {
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ti_stop(sc);
}
}
sc->ti_if_flags = ifp->if_flags;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_UNLOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK(sc);
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ti_setmulti(sc);
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_UNLOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
case SIOCSIFMEDIA:
case SIOCGIFMEDIA:
error = ifmedia_ioctl(ifp, ifr, &sc->ifmedia, command);
break;
Have the driver advertise that it is capable of hardware checksums, and allow the user to control the setting. (checksums are enabled by default)
2001-09-18 18:40:22 +00:00
case SIOCSIFCAP:
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK(sc);
Have the driver advertise that it is capable of hardware checksums, and allow the user to control the setting. (checksums are enabled by default)
2001-09-18 18:40:22 +00:00
mask = ifr->ifr_reqcap ^ ifp->if_capenable;
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
if ((mask & IFCAP_TXCSUM) != 0 &&
(ifp->if_capabilities & IFCAP_TXCSUM) != 0) {
ifp->if_capenable ^= IFCAP_TXCSUM;
if ((ifp->if_capenable & IFCAP_TXCSUM) != 0)
ifp->if_hwassist |= TI_CSUM_FEATURES;
else
ifp->if_hwassist &= ~TI_CSUM_FEATURES;
}
if ((mask & IFCAP_RXCSUM) != 0 &&
(ifp->if_capabilities & IFCAP_RXCSUM) != 0)
ifp->if_capenable ^= IFCAP_RXCSUM;
if ((mask & IFCAP_VLAN_HWTAGGING) != 0 &&
(ifp->if_capabilities & IFCAP_VLAN_HWTAGGING) != 0)
ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
if ((mask & IFCAP_VLAN_HWCSUM) != 0 &&
(ifp->if_capabilities & IFCAP_VLAN_HWCSUM) != 0)
ifp->if_capenable ^= IFCAP_VLAN_HWCSUM;
if ((mask & (IFCAP_TXCSUM | IFCAP_RXCSUM |
IFCAP_VLAN_HWTAGGING)) != 0) {
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
ti_init_locked(sc);
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
}
Whitespace pass.
2004-07-05 22:46:28 +00:00
}
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_UNLOCK(sc);
Don't abuse if_hwassist and make sure enabling corresponding TX/RX checksum offloading and VLAN hardware tag insertion/stripping from the currently enabled hardware offloading capabilities. Previously if_hwassist, which was initialized to TX/RX checksum offloading, was blindly used to enable both TX and RX checksum offloading such that disabling either TX or RX checksum offloading was not possible. ti(4) controllers support TX/RX checksum offloading with VLAN tagging so announce TX/RX checksum offloading capability over VLAN to vlan(4). Make VLAN hardware tag insertion/stripping honors currently enabled interface capability instead of blindly enabling VLAN hardware tagging. This change allows disabling hardware support of VLAN tag. Because ti(4) supports VLAN oversized frames, make network stack know the capability by setting if_hdrlen. While I'm here, rewrite SIOCSIFCAP handler and make sure to reinitialize controller whenever TX/RX checksum offloading and VLAN hardware tagging option is changed. The requirement of controller reinitialization comes from the limitation of Tigon I/II firmware. Tigon I/II firmware requires all related RCBs should be reinitialized whenever any of its hardware offloading capabilities change. vlan(4) is also notified whenever the parent interface's capability changes such that it can correctly handle TX/RX checksum offloading based on parent interface's enabled offloading capabilities. RX checksum offloading handler was changed to make upper stack use controller computed partial checksum value. Previously, ti(4) just set the computed value for any frames(IPv4, IPv6) and the value was not used in upper stack because driver didn't set CSUM_DATA_VALID such that upper network stack had to recompute checksum of TCP/UDP packets. I have no idea how this was not noticed for a long time. With this change, upper network stack does not have to fully recompute the checksum such that calculating pseudo checksum based on partial checksum is sufficient to know whether received packet's checksum is correct or not. However, I don't know why ti(4) does not have controller compute pseudo checksum as controller has ability to do it. I'm just guessing enabling that feature could trigger a firmware bug or could be slower than computing it on host side so just leave it as it was. In order not to produce false positives, ti(4) now checks whether controller actually computed IP or TCP/UDP checksum by checking ti_flags field.
2011-11-04 22:53:52 +00:00
VLAN_CAPABILITIES(ifp);
Have the driver advertise that it is capable of hardware checksums, and allow the user to control the setting. (checksums are enabled by default)
2001-09-18 18:40:22 +00:00
break;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
default:
o track either_ifattach/ether_ifdetach API changes o use if_input for input packet processing o don't strip the Ethernet header for input packets o use BPF_* macros bpf tapping o call ether_ioctl to handle default ioctl case o track vlan changes Reviewed by: many Approved by: re
2002-11-14 23:49:09 +00:00
error = ether_ioctl(ifp, command, data);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
break;
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (error);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
ti_open(struct cdev *dev, int flags, int fmt, struct thread *td)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
{
struct ti_softc *sc;
Sanitize the code relating to the /dev/ti%d entries. In particular evict the evil vnode sniffing code and use destroy_dev() instead.
2003-10-10 20:35:28 +00:00
sc = dev->si_drv1;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (sc == NULL)
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENODEV);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
TI_LOCK(sc);
sc->ti_flags |= TI_FLAG_DEBUGING;
TI_UNLOCK(sc);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
ti_close(struct cdev *dev, int flag, int fmt, struct thread *td)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
{
struct ti_softc *sc;
Sanitize the code relating to the /dev/ti%d entries. In particular evict the evil vnode sniffing code and use destroy_dev() instead.
2003-10-10 20:35:28 +00:00
sc = dev->si_drv1;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (sc == NULL)
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENODEV);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
TI_LOCK(sc);
sc->ti_flags &= ~TI_FLAG_DEBUGING;
TI_UNLOCK(sc);
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (0);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
/*
* This ioctl routine goes along with the Tigon character device.
*/
Whitespace nits
2004-07-05 22:35:18 +00:00
static int
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
ti_ioctl2(struct cdev *dev, u_long cmd, caddr_t addr, int flag,
struct thread *td)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
{
struct ti_softc *sc;
style. No functional changes.
2011-11-04 18:28:10 +00:00
int error;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
Sanitize the code relating to the /dev/ti%d entries. In particular evict the evil vnode sniffing code and use destroy_dev() instead.
2003-10-10 20:35:28 +00:00
sc = dev->si_drv1;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (sc == NULL)
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (ENODEV);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
error = 0;
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
switch (cmd) {
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
case TIIOCGETSTATS:
{
struct ti_stats *outstats;
outstats = (struct ti_stats *)addr;
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
bus_dmamap_sync(sc->ti_cdata.ti_gib_tag,
sc->ti_cdata.ti_gib_map, BUS_DMASYNC_POSTREAD);
bcopy(&sc->ti_rdata.ti_info->ti_stats, outstats,
style. No functional changes.
2011-11-10 22:15:11 +00:00
sizeof(struct ti_stats));
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
break;
}
case TIIOCGETPARAMS:
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_params *params;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
params = (struct ti_params *)addr;
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
params->ti_stat_ticks = sc->ti_stat_ticks;
params->ti_rx_coal_ticks = sc->ti_rx_coal_ticks;
params->ti_tx_coal_ticks = sc->ti_tx_coal_ticks;
params->ti_rx_max_coal_bds = sc->ti_rx_max_coal_bds;
params->ti_tx_max_coal_bds = sc->ti_tx_max_coal_bds;
params->ti_tx_buf_ratio = sc->ti_tx_buf_ratio;
params->param_mask = TI_PARAM_ALL;
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
break;
}
case TIIOCSETPARAMS:
{
struct ti_params *params;
params = (struct ti_params *)addr;
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (params->param_mask & TI_PARAM_STAT_TICKS) {
sc->ti_stat_ticks = params->ti_stat_ticks;
CSR_WRITE_4(sc, TI_GCR_STAT_TICKS, sc->ti_stat_ticks);
}
if (params->param_mask & TI_PARAM_RX_COAL_TICKS) {
sc->ti_rx_coal_ticks = params->ti_rx_coal_ticks;
CSR_WRITE_4(sc, TI_GCR_RX_COAL_TICKS,
sc->ti_rx_coal_ticks);
}
if (params->param_mask & TI_PARAM_TX_COAL_TICKS) {
sc->ti_tx_coal_ticks = params->ti_tx_coal_ticks;
CSR_WRITE_4(sc, TI_GCR_TX_COAL_TICKS,
sc->ti_tx_coal_ticks);
}
if (params->param_mask & TI_PARAM_RX_COAL_BDS) {
sc->ti_rx_max_coal_bds = params->ti_rx_max_coal_bds;
CSR_WRITE_4(sc, TI_GCR_RX_MAX_COAL_BD,
sc->ti_rx_max_coal_bds);
}
if (params->param_mask & TI_PARAM_TX_COAL_BDS) {
sc->ti_tx_max_coal_bds = params->ti_tx_max_coal_bds;
CSR_WRITE_4(sc, TI_GCR_TX_MAX_COAL_BD,
sc->ti_tx_max_coal_bds);
}
if (params->param_mask & TI_PARAM_TX_BUF_RATIO) {
sc->ti_tx_buf_ratio = params->ti_tx_buf_ratio;
CSR_WRITE_4(sc, TI_GCR_TX_BUFFER_RATIO,
sc->ti_tx_buf_ratio);
}
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
break;
}
case TIIOCSETTRACE: {
style. No functional changes.
2011-11-10 22:15:11 +00:00
ti_trace_type trace_type;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
trace_type = *(ti_trace_type *)addr;
/*
* Set tracing to whatever the user asked for. Setting
* this register to 0 should have the effect of disabling
* tracing.
*/
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
TI_LOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
CSR_WRITE_4(sc, TI_GCR_NIC_TRACING, trace_type);
It's bad idea to allocate large memory, 4KB, from stack. Pre-allocate the memory in device attach time. While I'm here remove unnecessary reassignment of error variable as it was already initialized. Also added a missing driver lock in TIIOCSETTRACE handler.
2011-11-14 18:40:04 +00:00
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
break;
}
case TIIOCGETTRACE: {
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_trace_buf *trace_buf;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t trace_start, cur_trace_ptr, trace_len;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
trace_buf = (struct ti_trace_buf *)addr;
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
trace_start = CSR_READ_4(sc, TI_GCR_NICTRACE_START);
cur_trace_ptr = CSR_READ_4(sc, TI_GCR_NICTRACE_PTR);
trace_len = CSR_READ_4(sc, TI_GCR_NICTRACE_LEN);
#if 0
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
if_printf(sc->ti_ifp, "trace_start = %#x, cur_trace_ptr = %#x, "
"trace_len = %d\n", trace_start,
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
cur_trace_ptr, trace_len);
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
if_printf(sc->ti_ifp, "trace_buf->buf_len = %d\n",
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
trace_buf->buf_len);
#endif
error = ti_copy_mem(sc, trace_start, min(trace_len,
style. No functional changes.
2011-11-10 22:15:11 +00:00
trace_buf->buf_len), (caddr_t)trace_buf->buf, 1, 1);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (error == 0) {
trace_buf->fill_len = min(trace_len,
style. No functional changes.
2011-11-10 22:15:11 +00:00
trace_buf->buf_len);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (cur_trace_ptr < trace_start)
trace_buf->cur_trace_ptr =
style. No functional changes.
2011-11-10 22:15:11 +00:00
trace_start - cur_trace_ptr;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
else
trace_buf->cur_trace_ptr =
style. No functional changes.
2011-11-10 22:15:11 +00:00
cur_trace_ptr - trace_start;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
} else
trace_buf->fill_len = 0;
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
break;
}
/*
* For debugging, five ioctls are needed:
* ALT_ATTACH
* ALT_READ_TG_REG
* ALT_WRITE_TG_REG
* ALT_READ_TG_MEM
* ALT_WRITE_TG_MEM
*/
case ALT_ATTACH:
/*
Whitespace nits
2004-07-05 22:35:18 +00:00
* From what I can tell, Alteon's Solaris Tigon driver
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
* only has one character device, so you have to attach
* to the Tigon board you're interested in. This seems
* like a not-so-good way to do things, since unless you
* subsequently specify the unit number of the device
Fix some "in in" typos in comments. PR: 121490 Submitted by: Anatoly Borodin <anatoly.borodin@gmail.com> Approved by: rwatson (mentor), jkoshy MFC after: 3 days
2008-03-26 07:32:08 +00:00
* you're interested in every ioctl, you'll only be
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
* able to debug one board at a time.
*/
break;
case ALT_READ_TG_MEM:
case ALT_WRITE_TG_MEM:
{
struct tg_mem *mem_param;
s/u_intXX_t/uintXX_t/g
2011-11-04 19:12:07 +00:00
uint32_t sram_end, scratch_end;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
mem_param = (struct tg_mem *)addr;
if (sc->ti_hwrev == TI_HWREV_TIGON) {
sram_end = TI_END_SRAM_I;
scratch_end = TI_END_SCRATCH_I;
} else {
sram_end = TI_END_SRAM_II;
scratch_end = TI_END_SCRATCH_II;
}
/*
* For now, we'll only handle accessing regular SRAM,
* nothing else.
*/
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
style. No functional changes.
2011-11-10 22:15:11 +00:00
if (mem_param->tgAddr >= TI_BEG_SRAM &&
mem_param->tgAddr + mem_param->len <= sram_end) {
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* In this instance, we always copy to/from user
* space, so the user space argument is set to 1.
*/
error = ti_copy_mem(sc, mem_param->tgAddr,
style. No functional changes.
2011-11-10 22:15:11 +00:00
mem_param->len, mem_param->userAddr, 1,
cmd == ALT_READ_TG_MEM ? 1 : 0);
} else if (mem_param->tgAddr >= TI_BEG_SCRATCH &&
mem_param->tgAddr <= scratch_end) {
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
error = ti_copy_scratch(sc, mem_param->tgAddr,
style. No functional changes.
2011-11-10 22:15:11 +00:00
mem_param->len, mem_param->userAddr, 1,
cmd == ALT_READ_TG_MEM ? 1 : 0, TI_PROCESSOR_A);
} else if (mem_param->tgAddr >= TI_BEG_SCRATCH_B_DEBUG &&
mem_param->tgAddr <= TI_BEG_SCRATCH_B_DEBUG) {
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (sc->ti_hwrev == TI_HWREV_TIGON) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
if_printf(sc->ti_ifp,
"invalid memory range for Tigon I\n");
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
error = EINVAL;
break;
}
Whitespace nits
2004-07-05 22:35:18 +00:00
error = ti_copy_scratch(sc, mem_param->tgAddr -
style. No functional changes.
2011-11-10 22:15:11 +00:00
TI_SCRATCH_DEBUG_OFF, mem_param->len,
mem_param->userAddr, 1,
cmd == ALT_READ_TG_MEM ? 1 : 0, TI_PROCESSOR_B);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
} else {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
if_printf(sc->ti_ifp, "memory address %#x len %d is "
"out of supported range\n",
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
mem_param->tgAddr, mem_param->len);
error = EINVAL;
}
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
break;
}
case ALT_READ_TG_REG:
case ALT_WRITE_TG_REG:
{
style. No functional changes.
2011-11-10 22:15:11 +00:00
struct tg_reg *regs;
uint32_t tmpval;
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
regs = (struct tg_reg *)addr;
/*
* Make sure the address in question isn't out of range.
*/
if (regs->addr > TI_REG_MAX) {
error = EINVAL;
break;
}
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_LOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
if (cmd == ALT_READ_TG_REG) {
bus_space_read_region_4(sc->ti_btag, sc->ti_bhandle,
style. No functional changes.
2011-11-10 22:15:11 +00:00
regs->addr, &tmpval, 1);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
regs->data = ntohl(tmpval);
#if 0
if ((regs->addr == TI_CPU_STATE)
|| (regs->addr == TI_CPU_CTL_B)) {
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
if_printf(sc->ti_ifp, "register %#x = %#x\n",
regs->addr, tmpval);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
#endif
} else {
tmpval = htonl(regs->data);
bus_space_write_region_4(sc->ti_btag, sc->ti_bhandle,
style. No functional changes.
2011-11-10 22:15:11 +00:00
regs->addr, &tmpval, 1);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
if_ti has been operating with locks for a while, so remove the GIANT markers. Also fix man potential locking problems in the cdev ioctl handler.
2005-12-10 01:25:46 +00:00
TI_UNLOCK(sc);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
break;
}
default:
error = ENOTTY;
break;
}
style(9): - Space before bracketized non-void function returns. - Space before condition for conditional blocks.
2004-07-05 22:42:07 +00:00
return (error);
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
}
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
ti_watchdog(void *arg)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
struct ifnet *ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
sc = arg;
TI_LOCK_ASSERT(sc);
callout_reset(&sc->ti_watchdog, hz, ti_watchdog, sc);
if (sc->ti_timer == 0 || --sc->ti_timer > 0)
return;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
/*
* When we're debugging, the chip is often stopped for long periods
* of time, and that would normally cause the watchdog timer to fire.
* Since that impedes debugging, we don't want to do that.
*/
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
if (sc->ti_flags & TI_FLAG_DEBUGING)
At long last, commit the zero copy sockets code. MAKEDEV: Add MAKEDEV glue for the ti(4) device nodes. ti.4: Update the ti(4) man page to include information on the TI_JUMBO_HDRSPLIT and TI_PRIVATE_JUMBOS kernel options, and also include information about the new character device interface and the associated ioctls. man9/Makefile: Add jumbo.9 and zero_copy.9 man pages and associated links. jumbo.9: New man page describing the jumbo buffer allocator interface and operation. zero_copy.9: New man page describing the general characteristics of the zero copy send and receive code, and what an application author should do to take advantage of the zero copy functionality. NOTES: Add entries for ZERO_COPY_SOCKETS, TI_PRIVATE_JUMBOS, TI_JUMBO_HDRSPLIT, MSIZE, and MCLSHIFT. conf/files: Add uipc_jumbo.c and uipc_cow.c. conf/options: Add the 5 options mentioned above. kern_subr.c: Receive side zero copy implementation. This takes "disposable" pages attached to an mbuf, gives them to a user process, and then recycles the user's page. This is only active when ZERO_COPY_SOCKETS is turned on and the kern.ipc.zero_copy.receive sysctl variable is set to 1. uipc_cow.c: Send side zero copy functions. Takes a page written by the user and maps it copy on write and assigns it kernel virtual address space. Removes copy on write mapping once the buffer has been freed by the network stack. uipc_jumbo.c: Jumbo disposable page allocator code. This allocates (optionally) disposable pages for network drivers that want to give the user the option of doing zero copy receive. uipc_socket.c: Add kern.ipc.zero_copy.{send,receive} sysctls that are enabled if ZERO_COPY_SOCKETS is turned on. Add zero copy send support to sosend() -- pages get mapped into the kernel instead of getting copied if they meet size and alignment restrictions. uipc_syscalls.c:Un-staticize some of the sf* functions so that they can be used elsewhere. (uipc_cow.c) if_media.c: In the SIOCGIFMEDIA ioctl in ifmedia_ioctl(), avoid calling malloc() with M_WAITOK. Return an error if the M_NOWAIT malloc fails. The ti(4) driver and the wi(4) driver, at least, call this with a mutex held. This causes witness warnings for 'ifconfig -a' with a wi(4) or ti(4) board in the system. (I've only verified for ti(4)). ip_output.c: Fragment large datagrams so that each segment contains a multiple of PAGE_SIZE amount of data plus headers. This allows the receiver to potentially do page flipping on receives. if_ti.c: Add zero copy receive support to the ti(4) driver. If TI_PRIVATE_JUMBOS is not defined, it now uses the jumbo(9) buffer allocator for jumbo receive buffers. Add a new character device interface for the ti(4) driver for the new debugging interface. This allows (a patched version of) gdb to talk to the Tigon board and debug the firmware. There are also a few additional debugging ioctls available through this interface. Add header splitting support to the ti(4) driver. Tweak some of the default interrupt coalescing parameters to more useful defaults. Add hooks for supporting transmit flow control, but leave it turned off with a comment describing why it is turned off. if_tireg.h: Change the firmware rev to 12.4.11, since we're really at 12.4.11 plus fixes from 12.4.13. Add defines needed for debugging. Remove the ti_stats structure, it is now defined in sys/tiio.h. ti_fw.h: 12.4.11 firmware. ti_fw2.h: 12.4.11 firmware, plus selected fixes from 12.4.13, and my header splitting patches. Revision 12.4.13 doesn't handle 10/100 negotiation properly. (This firmware is the same as what was in the tree previously, with the addition of header splitting support.) sys/jumbo.h: Jumbo buffer allocator interface. sys/mbuf.h: Add a new external mbuf type, EXT_DISPOSABLE, to indicate that the payload buffer can be thrown away / flipped to a userland process. socketvar.h: Add prototype for socow_setup. tiio.h: ioctl interface to the character portion of the ti(4) driver, plus associated structure/type definitions. uio.h: Change prototype for uiomoveco() so that we'll know whether the source page is disposable. ufs_readwrite.c:Update for new prototype of uiomoveco(). vm_fault.c: In vm_fault(), check to see whether we need to do a page based copy on write fault. vm_object.c: Add a new function, vm_object_allocate_wait(). This does the same thing that vm_object allocate does, except that it gives the caller the opportunity to specify whether it should wait on the uma_zalloc() of the object structre. This allows vm objects to be allocated while holding a mutex. (Without generating WITNESS warnings.) vm_object_allocate() is implemented as a call to vm_object_allocate_wait() with the malloc flag set to M_WAITOK. vm_object.h: Add prototype for vm_object_allocate_wait(). vm_page.c: Add page-based copy on write setup, clear and fault routines. vm_page.h: Add page based COW function prototypes and variable in the vm_page structure. Many thanks to Drew Gallatin, who wrote the zero copy send and receive code, and to all the other folks who have tested and reviewed this code over the years.
2002-06-26 03:37:47 +00:00
return;
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
ifp = sc->ti_ifp;
Use if_printf() and device_printf().
2005-09-29 16:47:08 +00:00
if_printf(ifp, "watchdog timeout -- resetting\n");
o Remove unnecessary controller reinitialization. o Do not blindly UP controller when MTU is changed. Reinitialize controller only if driver is running. o Remove useless ti_stop() in ti_watchdog() since ti_init_locked() always invokes ti_stop().
2011-11-07 18:26:38 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
ti_init_locked(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ifp->if_oerrors++;
}
/*
* Stop the adapter and free any mbufs allocated to the
* RX and TX lists.
*/
style: put return types on a line by themselves.
2002-08-24 00:02:03 +00:00
static void
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_stop(struct ti_softc *sc)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ifnet *ifp;
struct ti_cmd_desc cmd;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Bring big-endian architecture support for ti(4). . remove unnecessay header files after Scott's bus_dma(9) commit. . remove global variable tis which was introduced at the time of zero_copy(9) changes. The variable tis was not used at all. The same applyes to ti_links in softc so axe it. . deregister variables. . axe ti_vhandle and switch to use explicit register access for accessing NIC local memory. Creates three variants of ti_mem to read/write NIC local memory(ti_mem_read, ti_mem_write) and clearing NIC local memory(ti_mem_zero). This greatly enhances code readability and have ti(4) drop using shared memory scheme for Tigon 1. As Tigon 1 switched to use explicit register access for Tx, axe ti_tx_ring_nic/ti_cmd_ring in softc.(Tigon 2 used to host ring scheme which means there is no need to access NIC local memory via register access for Tx and NIC would DMA the modified Tx rings into its local memory.) [1] . introduce new macro TI_EVENT_*/TI_CMD_* to handle NIC envent/command. Instead of using bit fields assginment for accessing the event, use shift operations to set/get it. [1] . add additional check for valid DMA tags in ti_free_dmamaps(). . add missing bus_dmamap_sync/bus_dmamap_unload in ti_free_*_ring_*. . fix locking nits(MTX_RECURSE mutex) and make ti(4) MPSAFE. . change data type of ti_rdata_phys to bus_addr_t and don't blindly cast to uint32_t. . rearrange detach path and make ti(4) survive during device detach. . for Tigon 1, use explicit register access for checking Tx descriptors in ti_encap()/ti_txeof(). [1] . properly call bus_dmamap_sync(9) for updating statistics. . remove extra semicolon in ti_encap() . rewrite loading MAC address to work on strict-alignment architectures. . move TI_RD_OFF macro to if_tireg.h . axe ETHER_ALIGN as it's already defined in <net/ethernet.h>. . make macros immuine from expansion by adding parenthesis and do-while. . remove alpha specific hack as vtophys(9) is no longer used in ti(4) after Scott's bus_dma(9) fix. Reviewed by: scottl Obtained from: OpenBSD [1]
2005-12-28 02:57:19 +00:00
TI_LOCK_ASSERT(sc);
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ti_ifp;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Disable host interrupts. */
CSR_WRITE_4(sc, TI_MB_HOSTINTR, 1);
/*
* Tell firmware we're shutting down.
*/
TI_DO_CMD(TI_CMD_HOST_STATE, TI_CMD_CODE_STACK_DOWN, 0);
/* Halt and reinitialize. */
If ti_chipinit() fails in ti_stop(), ignore the error and release all allocated TX/RX buffer resources. If the interface is brought to up again after the error, we will leak allocated TX/RX buffers.
2011-11-07 22:53:06 +00:00
if (ti_chipinit(sc) == 0) {
ti_mem_zero(sc, 0x2000, 0x100000 - 0x2000);
/* XXX ignore init errors. */
ti_chipinit(sc);
}
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
/* Free the RX lists. */
ti_free_rx_ring_std(sc);
/* Free jumbo RX list. */
ti_free_rx_ring_jumbo(sc);
/* Free mini RX list. */
ti_free_rx_ring_mini(sc);
/* Free TX buffers. */
ti_free_tx_ring(sc);
sc->ti_ev_prodidx.ti_idx = 0;
sc->ti_return_prodidx.ti_idx = 0;
sc->ti_tx_considx.ti_idx = 0;
sc->ti_tx_saved_considx = TI_TXCONS_UNSET;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
- Add a private timer to drive the transmit watchdog instead of using if_watchdog and if_timer. - Fix some issues in detach for sn(4), ste(4), and ti(4). Primarily this means calling ether_ifdetach() before anything else.
2009-11-19 22:06:40 +00:00
callout_stop(&sc->ti_watchdog);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
/*
* Stop all chip I/O so that the kernel's probe routines don't
* get confused by errant DMAs when rebooting.
*/
Fix function prototype for device_shutdown method.
2007-11-22 02:45:00 +00:00
static int
Use ANSI function definations.
2011-11-04 17:07:53 +00:00
ti_shutdown(device_t dev)
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
{
style. No functional changes.
2011-11-04 18:28:10 +00:00
struct ti_softc *sc;
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
Convert the Alteon Tigon gigabit ethernet driver to newbus. Also upgrade to the latest firmware release from Alteon (12.3.12).
1999-07-23 02:10:11 +00:00
sc = device_get_softc(dev);
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
TI_LOCK(sc);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
ti_chipinit(sc);
First round of converting network drivers from spls to mutexes. This takes care of all the 10/100 and gigE PCI drivers that I've done. Next will be the wireless drivers, then the USB ones. I may pick up some stragglers along the way. I'm sort of playing this by ear: if anyone spots any places where I've screwed up horribly, please let me know.
2000-10-13 17:54:19 +00:00
TI_UNLOCK(sc);
Fix function prototype for device_shutdown method.
2007-11-22 02:45:00 +00:00
return (0);
Add driver support for gigabit ethernet adapters based on the Alteon Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed gigabit ethernet adapters out there which use the Alteon chipset so this driver covers a fair amount of hardware. I know that it works with the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and and Sun boards. The Netgear board is the cheapest (~$350US) but still yields fairly good performance. Support is provided for jumbo frames with all adapters (just set the MTU to something larger than 1500 bytes), as well as hardware multicast filtering and vlan tagging (in conjunction with the vlan support in -current, which I should merge into -stable soon). There are some hooks for checksum offload support, but they're turned off for now since FreeBSD doesn't have an officially sanctioned way to support checksum offloading (yet). I have not added the 'device ti0' entry to GENERIC since the driver with all the firmware compiled in is quite large, and it doesn't really fit into the category of generic hardware.
1999-04-06 17:08:31 +00:00
}
Export sysctl node for various interrupt moderation parameters and have administrators control them. ti(4) provides a character device to control various other features of driver via ioctls but users had to write their own code to manipulate these parameters. It seems some default values for these parameters are not optimal on today's system but leave it as it was and let administrators change them. The following parameters could be changed: dev.ti.%d.rx_coal_ticks dev.ti.%d.rx_max_coal_bds dev.ti.%d.tx_coal_ticks dev.ti.%d.tx_max_coal_bds dev.ti.%d.tx_buf_ratio dev.ti.%d.stat_ticks The interface has to be brought down and up again before a change takes effect. ti(4) controller supports hardware MAC counters with additional DMA statistics. So it's doable to export these counters via sysctl interface. Unfortunately, these counters are cumulative such that driver have to either send an explicit clear command to controller after extracting them or have to maintain internal counters to get actual changes. Neither look good to me so counters were not exported via sysctl.
2011-11-14 19:10:20 +00:00
static void
ti_sysctl_node(struct ti_softc *sc)
{
struct sysctl_ctx_list *ctx;
struct sysctl_oid_list *child;
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
char tname[32];
Export sysctl node for various interrupt moderation parameters and have administrators control them. ti(4) provides a character device to control various other features of driver via ioctls but users had to write their own code to manipulate these parameters. It seems some default values for these parameters are not optimal on today's system but leave it as it was and let administrators change them. The following parameters could be changed: dev.ti.%d.rx_coal_ticks dev.ti.%d.rx_max_coal_bds dev.ti.%d.tx_coal_ticks dev.ti.%d.tx_max_coal_bds dev.ti.%d.tx_buf_ratio dev.ti.%d.stat_ticks The interface has to be brought down and up again before a change takes effect. ti(4) controller supports hardware MAC counters with additional DMA statistics. So it's doable to export these counters via sysctl interface. Unfortunately, these counters are cumulative such that driver have to either send an explicit clear command to controller after extracting them or have to maintain internal counters to get actual changes. Neither look good to me so counters were not exported via sysctl.
2011-11-14 19:10:20 +00:00
ctx = device_get_sysctl_ctx(sc->ti_dev);
child = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->ti_dev));
Overhaul bus_dma(9) usage in driver: - Don't use a single big DMA block for all rings. Create separate DMA area for each ring instead. Currently the following DMA areas are created: Event ring, standard RX ring, jumbo RX ring, RX return ring, hardware MAC statistics and producer/consumer status area. For Tigon II, mini RX ring and TX ring are additionally created. - Added missing bus_dmamap_sync(9) in various TX/RX paths. - TX ring is no longer created for Tigon 1 such that it saves more resources on Tigon 1. - Data sheet is not clear about alignment requirement of each ring so use 32 bytes alignment for normal DMA area but use 64 bytes alignment for jumbo RX ring where the extended RX descriptor size is 64 bytes. - For each TX/RX buffers use separate DMA tag(e.g. the size of a DMA segment, total size of DMA segments etc). - Tigon allows separate DMA area for event producer, RX return producer and TX consumer which is really cool feature. This means TX and RX path could be independently run in parallel. However ti(4) uses a single driver lock so it's meaningless to have separate DMA area for these producer/consumer such that this change creates a single status DMA area. - It seems Tigon has no limits on DMA address space and I also don't see any problem with that but old comments in driver indicates there could be issues on descriptors being located in 64bit region. Introduce a tunable, dev.ti.%d.dac, to disable using 64bit DMA in driver. The default is 0 which means it would use full 64bit DMA. If there are DMA issues, users can disable it by setting the tunable to 0. - Do not increase watchdog timer in ti_txeof(). Previously driver increased the watchdog timer whenever there are queued TX frames. - When stat ticks is set to 0, skip processing ti_stats_update(), avoiding bus_dmamap_sync(9) and updating if_collisions counter. - MTU does not include FCS bytes, replace it with ETHER_VLAN_ENCAP_LEN. With these changes, ti(4) should work on PAE environments. Many thanks to Jay Borkenhagen for remote hardware access.
2011-11-14 20:38:14 +00:00
/* Use DAC */
sc->ti_dac = 1;
snprintf(tname, sizeof(tname), "dev.ti.%d.dac",
device_get_unit(sc->ti_dev));
TUNABLE_INT_FETCH(tname, &sc->ti_dac);
Export sysctl node for various interrupt moderation parameters and have administrators control them. ti(4) provides a character device to control various other features of driver via ioctls but users had to write their own code to manipulate these parameters. It seems some default values for these parameters are not optimal on today's system but leave it as it was and let administrators change them. The following parameters could be changed: dev.ti.%d.rx_coal_ticks dev.ti.%d.rx_max_coal_bds dev.ti.%d.tx_coal_ticks dev.ti.%d.tx_max_coal_bds dev.ti.%d.tx_buf_ratio dev.ti.%d.stat_ticks The interface has to be brought down and up again before a change takes effect. ti(4) controller supports hardware MAC counters with additional DMA statistics. So it's doable to export these counters via sysctl interface. Unfortunately, these counters are cumulative such that driver have to either send an explicit clear command to controller after extracting them or have to maintain internal counters to get actual changes. Neither look good to me so counters were not exported via sysctl.
2011-11-14 19:10:20 +00:00
SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "rx_coal_ticks", CTLFLAG_RW,
&sc->ti_rx_coal_ticks, 0, "Receive coalcesced ticks");
SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "rx_max_coal_bds", CTLFLAG_RW,
&sc->ti_rx_max_coal_bds, 0, "Receive max coalcesced BDs");
SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "tx_coal_ticks", CTLFLAG_RW,
&sc->ti_tx_coal_ticks, 0, "Send coalcesced ticks");
SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "tx_max_coal_bds", CTLFLAG_RW,
&sc->ti_tx_max_coal_bds, 0, "Send max coalcesced BDs");
SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "tx_buf_ratio", CTLFLAG_RW,
&sc->ti_tx_buf_ratio, 0,
"Ratio of NIC memory devoted to TX buffer");
SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "stat_ticks", CTLFLAG_RW,
&sc->ti_stat_ticks, 0,
"Number of clock ticks for statistics update interval");
/* Pull in device tunables. */
sc->ti_rx_coal_ticks = 170;
resource_int_value(device_get_name(sc->ti_dev),
device_get_unit(sc->ti_dev), "rx_coal_ticks",
&sc->ti_rx_coal_ticks);
sc->ti_rx_max_coal_bds = 64;
resource_int_value(device_get_name(sc->ti_dev),
device_get_unit(sc->ti_dev), "rx_max_coal_bds",
&sc->ti_rx_max_coal_bds);
sc->ti_tx_coal_ticks = TI_TICKS_PER_SEC / 500;
resource_int_value(device_get_name(sc->ti_dev),
device_get_unit(sc->ti_dev), "tx_coal_ticks",
&sc->ti_tx_coal_ticks);
sc->ti_tx_max_coal_bds = 32;
resource_int_value(device_get_name(sc->ti_dev),
device_get_unit(sc->ti_dev), "tx_max_coal_bds",
&sc->ti_tx_max_coal_bds);
sc->ti_tx_buf_ratio = 21;
resource_int_value(device_get_name(sc->ti_dev),
device_get_unit(sc->ti_dev), "tx_buf_ratio",
&sc->ti_tx_buf_ratio);
sc->ti_stat_ticks = 2 * TI_TICKS_PER_SEC;
resource_int_value(device_get_name(sc->ti_dev),
device_get_unit(sc->ti_dev), "stat_ticks",
&sc->ti_stat_ticks);
}
Reference in New Issue Copy Permalink