d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
fe182ba1d0
freebsd-dev/sys/dev/e1000/if_em.c

4539 lines
141 KiB
C
Raw Normal View History

Add copywrite notices, 2-clause BSD. Reported by: jmallett
2017-01-10 04:50:26 +00:00
/*-
* Copyright (c) 2016 Matt Macy <mmacy@nextbsd.org>
* 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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR 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 THE AUTHOR OR CONTRIBUTORS 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.
*/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* $FreeBSD$ */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
#include "if_em.h"
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
#include <sys/sbuf.h>
#include <machine/_inttypes.h>
#define em_mac_min e1000_82547
#define igb_mac_min e1000_82575
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
/*********************************************************************
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
* Driver version:
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*********************************************************************/
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
char em_driver_version[] = "7.6.1-k";
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
/*********************************************************************
* PCI Device ID Table
*
* Used by probe to select devices to load on
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
* Last field stores an index into e1000_strings
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
* Last entry must be all 0s
*
* { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
*********************************************************************/
Added support for 82541 and 82547 based adapters. - These have Intel gigabit PHY - 82547 uses CSA interface MFC after: 1 week
2003-03-21 21:47:31 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static pci_vendor_info_t em_vendor_info_array[] =
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Intel(R) PRO/1000 Network Connection - Legacy em*/
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Intel(R) PRO/1000 Network Connection - em */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 Network Connection"),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 Network Connection"),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) PRO/1000 Network Connection"),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) PRO/1000 Network Connection"),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) PRO/1000 Network Connection"),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) PRO/1000 Network Connection"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) PRO/1000 Network Connection"),
e1000: Add support for Kaby Lake generation i219 (4) and i219 (5) devices MFC after: 1 week Sponsored by: Intel Corporation
2017-01-19 18:52:38 +00:00
PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) PRO/1000 Network Connection"),
PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) PRO/1000 Network Connection"),
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
/* required last entry */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID_END
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
};
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static pci_vendor_info_t igb_vendor_info_array[] =
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* Intel(R) PRO/1000 Network Connection - igb */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
/* required last entry */
PVID_END
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
};
/*********************************************************************
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
* Function prototypes
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*********************************************************************/
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static void *em_register(device_t dev);
static void *igb_register(device_t dev);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int em_if_attach_pre(if_ctx_t ctx);
static int em_if_attach_post(if_ctx_t ctx);
static int em_if_detach(if_ctx_t ctx);
static int em_if_shutdown(if_ctx_t ctx);
static int em_if_suspend(if_ctx_t ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_if_resume(if_ctx_t ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
static int em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
static void em_if_queues_free(if_ctx_t ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static uint64_t em_if_get_counter(if_ctx_t, ift_counter);
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
static void em_if_init(if_ctx_t ctx);
static void em_if_stop(if_ctx_t ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void em_if_media_status(if_ctx_t, struct ifmediareq *);
static int em_if_media_change(if_ctx_t ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
static void em_if_timer(if_ctx_t ctx, uint16_t qid);
static void em_if_vlan_register(if_ctx_t ctx, u16 vtag);
static void em_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void em_identify_hardware(if_ctx_t ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_allocate_pci_resources(if_ctx_t ctx);
static void em_free_pci_resources(if_ctx_t ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void em_reset(if_ctx_t ctx);
static int em_setup_interface(if_ctx_t ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_setup_msix(if_ctx_t ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void em_initialize_transmit_unit(if_ctx_t ctx);
static void em_initialize_receive_unit(if_ctx_t ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static void em_if_enable_intr(if_ctx_t ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void em_if_disable_intr(if_ctx_t ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
static int em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
static void em_if_multi_set(if_ctx_t ctx);
static void em_if_update_admin_status(if_ctx_t ctx);
static void em_if_debug(if_ctx_t ctx);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
static void em_update_stats_counters(struct adapter *);
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
static void em_add_hw_stats(struct adapter *adapter);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_if_set_promisc(if_ctx_t ctx, int flags);
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
static void em_setup_vlan_hw_support(struct adapter *);
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
static int em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
static void em_print_nvm_info(struct adapter *);
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
static int em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_get_rs(SYSCTL_HANDLER_ARGS);
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
static void em_print_debug_info(struct adapter *);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
static int em_is_valid_ether_addr(u8 *);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
static int em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
static void em_add_int_delay_sysctl(struct adapter *, const char *,
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
const char *, struct em_int_delay_info *, int, int);
/* Management and WOL Support */
static void em_init_manageability(struct adapter *);
static void em_release_manageability(struct adapter *);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static void em_get_hw_control(struct adapter *);
static void em_release_hw_control(struct adapter *);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void em_get_wakeup(if_ctx_t ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static void em_enable_wakeup(if_ctx_t ctx);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
static int em_enable_phy_wakeup(struct adapter *);
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
static void em_disable_aspm(struct adapter *);
The procedure of raceless switching between polling mode and taskqueued interrupt mode is going to be quite complex. Since the polling mode is considered legacy feature for em(4) driver, the decision is made to make polling and new interrupt handler mutually exclusive, selected at compile time. If kernel is compiled with DEVICE_POLLING, the fast taskqueued interrupt handler code is disabled and the em_poll() and legacy em_intr() functions are enabled. Otherwise, legacy functions are disabled and only em_intr_fast() code is compiled. Discussed with: scottl
2006-06-06 08:03:49 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
int em_intr(void *arg);
static void em_disable_promisc(if_ctx_t ctx);
This delta syncs the em and igb drivers with Intel, adds header split and SCTP support into the igb driver. Various small improvements and fixes. MFC after: 2 weeks
2009-04-10 00:05:46 +00:00
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
/* MSIX handlers */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_if_msix_intr_assign(if_ctx_t, int);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int em_msix_link(void *);
static void em_handle_link(void *context);
static void em_enable_vectors_82574(if_ctx_t);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
static int em_set_flowcntl(SYSCTL_HANDLER_ARGS);
Change the interface to the Energy Efficient Ethernet (EEE) setting in the igb and em driver. This was necessitated by a shared code change that I was given late in the game, a data type changed from bool to int, in the last update I dealt with it by a cast, but it was pointed out (thanks jhb) that there was a potential problem with this. John suggested this safer approach, and it is fine with me... MFC after:2 days (to catch the 9.1 update)
2012-07-07 20:21:05 +00:00
static int em_sysctl_eee(SYSCTL_HANDLER_ARGS);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static void em_if_led_func(if_ctx_t ctx, int onoff);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int em_get_regs(SYSCTL_HANDLER_ARGS);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void lem_smartspeed(struct adapter *adapter);
static void igb_configure_queues(struct adapter *adapter);
Tighten up the rx mbuf refresh code, there were some discrepencies from the igb version which was the target. Change the message when neither MSI or MSIX are enabled and a fallback to Legacy interrupts happen, the existing message was confusing.
2010-09-07 20:13:08 +00:00
Take out em_poll() prototype from under EM_FAST_IRQ control. Reported by: tindebox compiling a LINT kernel
2007-11-21 12:55:33 +00:00
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
/*********************************************************************
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
* FreeBSD Device Interface Entry Points
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*********************************************************************/
static device_method_t em_methods[] = {
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Device interface */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
DEVMETHOD(device_register, em_register),
DEVMETHOD(device_probe, iflib_device_probe),
DEVMETHOD(device_attach, iflib_device_attach),
DEVMETHOD(device_detach, iflib_device_detach),
DEVMETHOD(device_shutdown, iflib_device_shutdown),
DEVMETHOD(device_suspend, iflib_device_suspend),
DEVMETHOD(device_resume, iflib_device_resume),
DEVMETHOD_END
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
};
static device_method_t igb_methods[] = {
/* Device interface */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
DEVMETHOD(device_register, igb_register),
DEVMETHOD(device_probe, iflib_device_probe),
DEVMETHOD(device_attach, iflib_device_attach),
DEVMETHOD(device_detach, iflib_device_detach),
DEVMETHOD(device_shutdown, iflib_device_shutdown),
DEVMETHOD(device_suspend, iflib_device_suspend),
DEVMETHOD(device_resume, iflib_device_resume),
DEVMETHOD_END
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
};
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
static driver_t em_driver = {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
"em", em_methods, sizeof(struct adapter),
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
};
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static devclass_t em_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(em, pci, em_driver, em_devclass, 0, 0);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
- Express hard dependencies on bus (pci, isa, pccard) and network layer (ether). - Don't abuse module names to facilitate ifconfig module loading; such abuse isn't really needed. (And if we do need type information associated with a module then we should make it explicit and not use hacks.)
2003-04-15 06:37:30 +00:00
MODULE_DEPEND(em, pci, 1, 1, 1);
MODULE_DEPEND(em, ether, 1, 1, 1);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
MODULE_DEPEND(em, iflib, 1, 1, 1);
static driver_t igb_driver = {
"igb", igb_methods, sizeof(struct adapter),
};
static devclass_t igb_devclass;
DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
MODULE_DEPEND(igb, pci, 1, 1, 1);
MODULE_DEPEND(igb, ether, 1, 1, 1);
MODULE_DEPEND(igb, iflib, 1, 1, 1);
static device_method_t em_if_methods[] = {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
DEVMETHOD(ifdi_attach_post, em_if_attach_post),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
DEVMETHOD(ifdi_detach, em_if_detach),
DEVMETHOD(ifdi_shutdown, em_if_shutdown),
DEVMETHOD(ifdi_suspend, em_if_suspend),
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
DEVMETHOD(ifdi_resume, em_if_resume),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
DEVMETHOD(ifdi_init, em_if_init),
DEVMETHOD(ifdi_stop, em_if_stop),
DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
DEVMETHOD(ifdi_intr_enable, em_if_enable_intr),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
DEVMETHOD(ifdi_intr_disable, em_if_disable_intr),
DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
DEVMETHOD(ifdi_queues_free, em_if_queues_free),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
DEVMETHOD(ifdi_multi_set, em_if_multi_set),
DEVMETHOD(ifdi_media_status, em_if_media_status),
DEVMETHOD(ifdi_media_change, em_if_media_change),
DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
DEVMETHOD(ifdi_timer, em_if_timer),
DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
DEVMETHOD(ifdi_get_counter, em_if_get_counter),
DEVMETHOD(ifdi_led_func, em_if_led_func),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
DEVMETHOD(ifdi_debug, em_if_debug),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
DEVMETHOD_END
};
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/*
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
* note that if (adapter->msix_mem) is replaced by:
* if (adapter->intr_type == IFLIB_INTR_MSIX)
*/
static driver_t em_if_driver = {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
"em_if", em_if_methods, sizeof(struct adapter)
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
};
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
/*********************************************************************
* Tunable default values.
*********************************************************************/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
#define EM_TICKS_TO_USECS(ticks) ((1024 * (ticks) + 500) / 1000)
#define EM_USECS_TO_TICKS(usecs) ((1000 * (usecs) + 512) / 1024)
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
#define M_TSO_LEN 66
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
if_lem.c: make sure that lem_rxeof() can drain the entire rx queue irrespective of the setting of lem_rx_process_limit, while giving a chance to the taskqueue scheduler to act after each chunk. This makes lem_rxeof similar to the one in if_em.c and if_igb.c . if_lem.c and if_em.c: add a sysctl to manually configure the 'itr' moderation register. Approved by: Jack Vogel
2013-05-09 17:07:30 +00:00
#define MAX_INTS_PER_SEC 8000
#define DEFAULT_ITR (1000000000/(MAX_INTS_PER_SEC * 256))
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Allow common code without TSO */
#ifndef CSUM_TSO
#define CSUM_TSO 0
#endif
Increase EM_MAX_SCATTER to 64 such that the size of em_xmit()::segs[EM_MAX_SCATTER] doesn't get overrun by things like NFS that can and do shove more than 32 segs when being used with em(4) and TSO4. Update tso handling code in em_xmit() with update from jhb@ in email thread: https://lists.freebsd.org/pipermail/freebsd-net/2014-July/039306.html set ifp->if_hw_tsomax, ifp->if_hw_tsomaxsegcount & ifp->if_hw_tsomaxsegsize to appropriate values. Define a TSO workaround "magic" number of 4 that is used to avoid an alignment issue in hardware. Change a couple of integer values that were used as booleans to actual bool types. Ensure that em_enable_intr() enables the appropriate mask of interrupts and not just a hardcoded define of values. PR: 200221 199174 195078 Differential Revision: https://reviews.freebsd.org/D3192 Reviewed by: erj jhb hiren MFC after: 2 weeks Sponsored by: Limelight Networks
2015-08-16 19:43:44 +00:00
#define TSO_WORKAROUND 4
Mark all SYSCTL_NODEs static that have no corresponding SYSCTL_DECLs. The SYSCTL_NODE macro defines a list that stores all child-elements of that node. If there's no SYSCTL_DECL macro anywhere else, there's no reason why it shouldn't be static.
2011-11-07 15:43:11 +00:00
static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
- Add read-only sysctls for all of the tunables supported by the igb and em drivers. - Make the per-instance 'enable_aim' sysctl truly per-instance by having it change a per-instance variable (which is used to control AIM) rather than having all of the per-instance sysctls operate on a single global variable. Reviewed by: jfv (earlier version) MFC after: 1 week
2011-06-29 16:20:52 +00:00
Add capability to disable CRC stripping. This breaks IPMI/BMC capabilities on certain adatpers. Linux has been doing the exact same thing since 2008 https://github.com/torvalds/linux/commit/eb7c3adb1ca92450870dbb0d347fc986cd5e2af4 PR: 161277 Differential Revision: https://reviews.freebsd.org/D3282 Submitted by: Fravadona@gmail.com Reviewed by: erj wblock MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks
2015-08-16 19:06:23 +00:00
static int em_disable_crc_stripping = 0;
SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
&em_disable_crc_stripping, 0, "Disable CRC Stripping");
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
- Add read-only sysctls for all of the tunables supported by the igb and em drivers. - Make the per-instance 'enable_aim' sysctl truly per-instance by having it change a per-instance variable (which is used to control AIM) rather than having all of the per-instance sysctls operate on a single global variable. Reviewed by: jfv (earlier version) MFC after: 1 week
2011-06-29 16:20:52 +00:00
SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
0, "Default transmit interrupt delay in usecs");
SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
0, "Default receive interrupt delay in usecs");
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
- Add read-only sysctls for all of the tunables supported by the igb and em drivers. - Make the per-instance 'enable_aim' sysctl truly per-instance by having it change a per-instance variable (which is used to control AIM) rather than having all of the per-instance sysctls operate on a single global variable. Reviewed by: jfv (earlier version) MFC after: 1 week
2011-06-29 16:20:52 +00:00
SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
&em_tx_abs_int_delay_dflt, 0,
"Default transmit interrupt delay limit in usecs");
SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
&em_rx_abs_int_delay_dflt, 0,
"Default receive interrupt delay limit in usecs");
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
static int em_smart_pwr_down = FALSE;
- Add read-only sysctls for all of the tunables supported by the igb and em drivers. - Make the per-instance 'enable_aim' sysctl truly per-instance by having it change a per-instance variable (which is used to control AIM) rather than having all of the per-instance sysctls operate on a single global variable. Reviewed by: jfv (earlier version) MFC after: 1 week
2011-06-29 16:20:52 +00:00
SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
0, "Set to true to leave smart power down enabled on newer adapters");
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Controls whether promiscuous also shows bad packets */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int em_debug_sbp = TRUE;
- Add read-only sysctls for all of the tunables supported by the igb and em drivers. - Make the per-instance 'enable_aim' sysctl truly per-instance by having it change a per-instance variable (which is used to control AIM) rather than having all of the per-instance sysctls operate on a single global variable. Reviewed by: jfv (earlier version) MFC after: 1 week
2011-06-29 16:20:52 +00:00
SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
"Show bad packets in promiscuous mode");
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* How many packets rxeof tries to clean at a time */
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
static int em_rx_process_limit = 100;
- Add read-only sysctls for all of the tunables supported by the igb and em drivers. - Make the per-instance 'enable_aim' sysctl truly per-instance by having it change a per-instance variable (which is used to control AIM) rather than having all of the per-instance sysctls operate on a single global variable. Reviewed by: jfv (earlier version) MFC after: 1 week
2011-06-29 16:20:52 +00:00
SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
&em_rx_process_limit, 0,
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
"Maximum number of received packets to process "
"at a time, -1 means unlimited");
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
/* Energy efficient ethernet - default to OFF */
Change the interface to the Energy Efficient Ethernet (EEE) setting in the igb and em driver. This was necessitated by a shared code change that I was given late in the game, a data type changed from bool to int, in the last update I dealt with it by a cast, but it was pointed out (thanks jhb) that there was a potential problem with this. John suggested this safer approach, and it is fine with me... MFC after:2 days (to catch the 9.1 update)
2012-07-07 20:21:05 +00:00
static int eee_setting = 1;
- Add read-only sysctls for all of the tunables supported by the igb and em drivers. - Make the per-instance 'enable_aim' sysctl truly per-instance by having it change a per-instance variable (which is used to control AIM) rather than having all of the per-instance sysctls operate on a single global variable. Reviewed by: jfv (earlier version) MFC after: 1 week
2011-06-29 16:20:52 +00:00
SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
"Enable Energy Efficient Ethernet");
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*
** Tuneable Interrupt rate
*/
static int em_max_interrupt_rate = 8000;
SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
&em_max_interrupt_rate, 0, "Maximum interrupts per second");
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Global used in WOL setup with multiport cards */
static int global_quad_port_a = 0;
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
extern struct if_txrx igb_txrx;
extern struct if_txrx em_txrx;
extern struct if_txrx lem_txrx;
static struct if_shared_ctx em_sctx_init = {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
.isc_magic = IFLIB_MAGIC,
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
.isc_q_align = PAGE_SIZE,
.isc_tx_maxsize = EM_TSO_SIZE,
.isc_tx_maxsegsize = PAGE_SIZE,
.isc_rx_maxsize = MJUM9BYTES,
.isc_rx_nsegments = 1,
.isc_rx_maxsegsize = MJUM9BYTES,
.isc_nfl = 1,
.isc_nrxqs = 1,
.isc_ntxqs = 1,
.isc_admin_intrcnt = 1,
.isc_vendor_info = em_vendor_info_array,
.isc_driver_version = em_driver_version,
.isc_driver = &em_if_driver,
Some small packet performance improvements If the packet is smaller than MTU, disable the TSO flags. Move TCP header parsing inside the IS_TSO?() test. Add a new IFLIB_NEED_ZERO_CSUM flag to indicate the checksums need to be zeroed before TX. Reviewed by: sbruno Approved by: sbruno (mentor) Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D12442
2017-09-23 01:33:20 +00:00
.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
.isc_nrxd_min = {EM_MIN_RXD},
.isc_ntxd_min = {EM_MIN_TXD},
.isc_nrxd_max = {EM_MAX_RXD},
.isc_ntxd_max = {EM_MAX_TXD},
.isc_nrxd_default = {EM_DEFAULT_RXD},
.isc_ntxd_default = {EM_DEFAULT_TXD},
};
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if_shared_ctx_t em_sctx = &em_sctx_init;
static struct if_shared_ctx igb_sctx_init = {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
.isc_magic = IFLIB_MAGIC,
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
.isc_q_align = PAGE_SIZE,
.isc_tx_maxsize = EM_TSO_SIZE,
.isc_tx_maxsegsize = PAGE_SIZE,
.isc_rx_maxsize = MJUM9BYTES,
.isc_rx_nsegments = 1,
.isc_rx_maxsegsize = MJUM9BYTES,
.isc_nfl = 1,
.isc_nrxqs = 1,
.isc_ntxqs = 1,
.isc_admin_intrcnt = 1,
.isc_vendor_info = igb_vendor_info_array,
.isc_driver_version = em_driver_version,
.isc_driver = &em_if_driver,
Some small packet performance improvements If the packet is smaller than MTU, disable the TSO flags. Move TCP header parsing inside the IS_TSO?() test. Add a new IFLIB_NEED_ZERO_CSUM flag to indicate the checksums need to be zeroed before TX. Reviewed by: sbruno Approved by: sbruno (mentor) Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D12442
2017-09-23 01:33:20 +00:00
.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
.isc_nrxd_min = {EM_MIN_RXD},
.isc_ntxd_min = {EM_MIN_TXD},
Add a different #define for the maximum number of transmit and recieve descriptors for the igb(4) class of devices. This will allow a better definition for maximum going forward. Some igb(4) devices support more than the default 4K. Reported by: Jason (j@nitrology.com) Sponsored by: Limelight Networks
2017-08-25 22:38:55 +00:00
.isc_nrxd_max = {IGB_MAX_RXD},
.isc_ntxd_max = {IGB_MAX_TXD},
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
.isc_nrxd_default = {EM_DEFAULT_RXD},
.isc_ntxd_default = {EM_DEFAULT_TXD},
};
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if_shared_ctx_t igb_sctx = &igb_sctx_init;
add netmap support for "em", "lem", "igb" and "re". On my hardware, "em" in netmap mode does about 1.388 Mpps on one card (on an Asus motherboard), and 1.1 Mpps on another card (PCIe bus). Both seem to be NIC-limited, because i have the same rate even with the CPU running at 150 MHz. On the "re" driver the tx throughput is around 420-450 Kpps on various (8111C and the like) chipsets. On the Rx side performance seems much better, and i can receive the full load generated by the "em" cards. "igb" is untested as i don't have the hardware.
2011-12-05 15:33:13 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*****************************************************************
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
* Dump Registers
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
****************************************************************/
#define IGB_REGS_LEN 739
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int em_get_regs(SYSCTL_HANDLER_ARGS)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = (struct adapter *)arg1;
struct e1000_hw *hw = &adapter->hw;
struct sbuf *sb;
Fix malloc() uses in em_get_regs(). Do not use malloc(M_NOWAIT), wait is possible there, and the malloc failures where not checked. Do not forget to free malloced memory. Reported and tested by: pho Approved by: sbruno Sponsored by: The FreeBSD Foundation
2017-09-08 14:54:07 +00:00
u32 *regs_buff;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
int rc;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Fix malloc() uses in em_get_regs(). Do not use malloc(M_NOWAIT), wait is possible there, and the malloc failures where not checked. Do not forget to free malloced memory. Reported and tested by: pho Approved by: sbruno Sponsored by: The FreeBSD Foundation
2017-09-08 14:54:07 +00:00
regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
rc = sysctl_wire_old_buffer(req, 0);
MPASS(rc == 0);
Fix malloc() uses in em_get_regs(). Do not use malloc(M_NOWAIT), wait is possible there, and the malloc failures where not checked. Do not forget to free malloced memory. Reported and tested by: pho Approved by: sbruno Sponsored by: The FreeBSD Foundation
2017-09-08 14:54:07 +00:00
if (rc != 0) {
free(regs_buff, M_DEVBUF);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (rc);
Fix malloc() uses in em_get_regs(). Do not use malloc(M_NOWAIT), wait is possible there, and the malloc failures where not checked. Do not forget to free malloced memory. Reported and tested by: pho Approved by: sbruno Sponsored by: The FreeBSD Foundation
2017-09-08 14:54:07 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
MPASS(sb != NULL);
Fix malloc() uses in em_get_regs(). Do not use malloc(M_NOWAIT), wait is possible there, and the malloc failures where not checked. Do not forget to free malloced memory. Reported and tested by: pho Approved by: sbruno Sponsored by: The FreeBSD Foundation
2017-09-08 14:54:07 +00:00
if (sb == NULL) {
free(regs_buff, M_DEVBUF);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
return (ENOMEM);
Fix malloc() uses in em_get_regs(). Do not use malloc(M_NOWAIT), wait is possible there, and the malloc failures where not checked. Do not forget to free malloced memory. Reported and tested by: pho Approved by: sbruno Sponsored by: The FreeBSD Foundation
2017-09-08 14:54:07 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* General Registers */
regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_printf(sb, "General Registers\n");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
sbuf_printf(sb, "Interrupt Registers\n");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_printf(sb, "RX Registers\n");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
sbuf_printf(sb, "TX Registers\n");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
Fix malloc() uses in em_get_regs(). Do not use malloc(M_NOWAIT), wait is possible there, and the malloc failures where not checked. Do not forget to free malloced memory. Reported and tested by: pho Approved by: sbruno Sponsored by: The FreeBSD Foundation
2017-09-08 14:54:07 +00:00
free(regs_buff, M_DEVBUF);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
#ifdef DUMP_DESCS
{
if_softc_ctx_t scctx = adapter->shared;
struct rx_ring *rxr = &rx_que->rxr;
struct tx_ring *txr = &tx_que->txr;
int ntxd = scctx->isc_ntxd[0];
int nrxd = scctx->isc_nrxd[0];
int j;
for (j = 0; j < nrxd; j++) {
u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
u32 length = le32toh(rxr->rx_base[j].wb.upper.length);
sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 " Error:%d Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
}
for (j = 0; j < min(ntxd, 256); j++) {
unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x eop: %d DD=%d\n",
j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
}
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
#endif
rc = sbuf_finish(sb);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
sbuf_delete(sb);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return(rc);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
static void *
em_register(device_t dev)
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (em_sctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
static void *
igb_register(device_t dev)
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (igb_sctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
static int
em_set_num_queues(if_ctx_t ctx)
{
struct adapter *adapter = iflib_get_softc(ctx);
int maxqueues;
/* Sanity check based on HW */
switch (adapter->hw.mac.type) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
case e1000_82576:
case e1000_82580:
case e1000_i350:
case e1000_i354:
maxqueues = 8;
break;
case e1000_i210:
case e1000_82575:
maxqueues = 4;
break;
case e1000_i211:
case e1000_82574:
maxqueues = 2;
break;
default:
maxqueues = 1;
break;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
return (maxqueues);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
#define EM_CAPS \
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | \
IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
#define IGB_CAPS \
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWCSUM | \
IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU | IFCAP_TXCSUM_IPV6 | IFCAP_HWCSUM_IPV6 | IFCAP_JUMBO_MTU;
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
/*********************************************************************
* Device initialization routine
*
* The attach entry point is called when the driver is being loaded.
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
* This routine identifies the type of hardware, allocates all resources
* and initializes the hardware.
*
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
* return 0 on success, positive on failure
*********************************************************************/
static int
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_if_attach_pre(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if_softc_ctx_t scctx;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
device_t dev;
struct e1000_hw *hw;
int error = 0;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
INIT_DEBUGOUT("em_if_attach_pre begin");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
dev = iflib_get_dev(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
adapter = iflib_get_softc(ctx);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
This isn't functionally identical. In some cases a hint to disable unit 0 would in fact disable all units. This reverts r241856 Approved by: cperciva (implicit)
2012-10-22 13:06:09 +00:00
if (resource_disabled("em", device_get_unit(dev))) {
device_printf(dev, "Disabled by device hint\n");
return (ENXIO);
}
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
adapter->ctx = ctx;
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
adapter->dev = adapter->osdep.dev = dev;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
scctx = adapter->shared = iflib_get_softc_ctx(ctx);
adapter->media = iflib_get_media(ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
hw = &adapter->hw;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
adapter->tx_process_limit = scctx->isc_ntxd[0];
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
/* SYSCTL stuff */
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
em_sysctl_nvm_info, "I", "NVM Information");
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
em_sysctl_debug_info, "I", "Debug Information");
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
em_set_flowcntl, "I", "Flow Control");
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "reg_dump", CTLTYPE_STRING | CTLFLAG_RD, adapter, 0,
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_get_regs, "A", "Dump Registers");
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "rs_dump", CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
em_get_rs, "I", "Dump RS indexes");
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Couple of changes, back down on last TSO change, instead make old adapter list still capable, but only PCI-E adapters are now enabled. The user can enable older PCI-X or PCI adapters using ifconfig. Secondly, Arthur Hartwig pointed out my MSI change was not working correctly, changed to something that now does. Thanks Arthur. There was also a fundamental bug in the 82575 MSIX code, the MSIX registers had to be mapped, opps :) Rubber-stamped by: Pdeuskar
2007-05-17 00:14:03 +00:00
/* Determine hardware and mac info */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_identify_hardware(ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* Set isc_msix_bar */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
scctx->isc_tx_nsegments = EM_MAX_SCATTER;
scctx->isc_tx_tso_segments_max = scctx->isc_tx_nsegments;
scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
device_printf(dev, "attach_pre capping queues at %d\n", scctx->isc_ntxqsets_max);
scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
if (adapter->hw.mac.type >= igb_mac_min) {
Attempt to use the "new" BAR address for newer igb(4) devices. This code was dropped during the IFLIB migration. Reported by: olivier Reviewed by: mmacy@nextbsd.org
2017-01-12 14:18:52 +00:00
int try_second_bar;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
scctx->isc_txrx = &igb_txrx;
scctx->isc_capenable = IGB_CAPS;
scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO | CSUM_IP6_TCP \
| CSUM_IP6_UDP | CSUM_IP6_TCP;
if (adapter->hw.mac.type != e1000_82575)
scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
Attempt to use the "new" BAR address for newer igb(4) devices. This code was dropped during the IFLIB migration. Reported by: olivier Reviewed by: mmacy@nextbsd.org
2017-01-12 14:18:52 +00:00
/*
** Some new devices, as with ixgbe, now may
** use a different BAR, so we need to keep
** track of which is used.
*/
try_second_bar = pci_read_config(dev, scctx->isc_msix_bar, 4);
if (try_second_bar == 0)
scctx->isc_msix_bar += 4;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
} else if (adapter->hw.mac.type >= em_mac_min) {
scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
scctx->isc_txrx = &em_txrx;
scctx->isc_capenable = EM_CAPS;
scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
} else {
scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
scctx->isc_txrx = &lem_txrx;
scctx->isc_capenable = EM_CAPS;
if (adapter->hw.mac.type < e1000_82543)
scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
scctx->isc_msix_bar = 0;
}
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Setup PCI resources */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (em_allocate_pci_resources(ctx)) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
device_printf(dev, "Allocation of PCI resources failed\n");
error = ENXIO;
goto err_pci;
}
/*
** For ICH8 and family we need to
** map the flash memory, and this
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
** must happen after the MAC is
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
** identified
*/
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
if ((hw->mac.type == e1000_ich8lan) ||
(hw->mac.type == e1000_ich9lan) ||
(hw->mac.type == e1000_ich10lan) ||
(hw->mac.type == e1000_pchlan) ||
Refresh on the shared code for the E1000 drivers. - bear with me, there are lots of white space changes, I would not do them, but I am a mere consumer of this stuff and if these drivers are to stay in shape they need to be taken. em driver changes: support for the new i217/i218 interfaces igb driver changes: - TX mq start has a quick turnaround to the stack - Link/media handling improvement - When link status changes happen the current flow control state will now be displayed. - A few white space/style changes. lem driver changes: - the shared code uncovered a bogus write to the RLPML register (which does not exist in this hardware) in the vlan code,this is removed.
2013-02-21 00:25:45 +00:00
(hw->mac.type == e1000_pch2lan) ||
(hw->mac.type == e1000_pch_lpt)) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
int rid = EM_BAR_TYPE_FLASH;
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
adapter->flash = bus_alloc_resource_any(dev,
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
SYS_RES_MEMORY, &rid, RF_ACTIVE);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
if (adapter->flash == NULL) {
device_printf(dev, "Mapping of Flash failed\n");
error = ENXIO;
goto err_pci;
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* This is used in the shared code */
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
hw->flash_address = (u8 *)adapter->flash;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->osdep.flash_bus_space_tag =
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
rman_get_bustag(adapter->flash);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->osdep.flash_bus_space_handle =
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
rman_get_bushandle(adapter->flash);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
}
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
/*
** In the new SPT device flash is not a
dev/e1000,ixgbe: minor spelling fixes. No functional change. Differential Revision: https://reviews.freebsd.org/D6177
2016-05-06 22:54:56 +00:00
** separate BAR, rather it is also in BAR0,
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
** so use the same tag and an offset handle for the
** FLASH read/write macros in the shared code.
*/
else if (hw->mac.type == e1000_pch_spt) {
adapter->osdep.flash_bus_space_tag =
adapter->osdep.mem_bus_space_tag;
adapter->osdep.flash_bus_space_handle =
adapter->osdep.mem_bus_space_handle
+ E1000_FLASH_BASE_ADDR;
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Do Shared Code initialization */
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
error = e1000_setup_init_funcs(hw, TRUE);
if (error) {
device_printf(dev, "Setup of Shared code failed, error %d\n",
error);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
error = ENXIO;
goto err_pci;
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_setup_msix(ctx);
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
e1000_get_bus_info(hw);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
/* Set up some sysctls for the tunable interrupt delays */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_add_int_delay_sysctl(adapter, "rx_int_delay",
"receive interrupt delay in usecs", &adapter->rx_int_delay,
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_add_int_delay_sysctl(adapter, "tx_int_delay",
"transmit interrupt delay in usecs", &adapter->tx_int_delay,
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
"receive interrupt delay limit in usecs",
&adapter->rx_abs_int_delay,
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
E1000_REGISTER(hw, E1000_RADV),
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
em_rx_abs_int_delay_dflt);
em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
"transmit interrupt delay limit in usecs",
&adapter->tx_abs_int_delay,
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
E1000_REGISTER(hw, E1000_TADV),
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
em_tx_abs_int_delay_dflt);
if_lem.c: make sure that lem_rxeof() can drain the entire rx queue irrespective of the setting of lem_rx_process_limit, while giving a chance to the taskqueue scheduler to act after each chunk. This makes lem_rxeof similar to the one in if_em.c and if_igb.c . if_lem.c and if_em.c: add a sysctl to manually configure the 'itr' moderation register. Approved by: Jack Vogel
2013-05-09 17:07:30 +00:00
em_add_int_delay_sysctl(adapter, "itr",
"interrupt delay limit in usecs/4",
&adapter->tx_itr,
E1000_REGISTER(hw, E1000_ITR),
DEFAULT_ITR);
- Backout last change, since it is memory overkill for a non busy host or for a notebook with em(4) adapter. - Introduce tunables em.hw.txd and em.hw.rxd, which allow administrator to configure number of transmit and receive descriptors. - Check em.hw.txd and em.hw.rxd against hardware limits [*] and require them to be multiple of 128. [*] According to comments in if_em.h the 82540EM/82541ER chips can handle more than 256 descriptors. Since we don't have this hardware to test, we decided to mimic NetBSD wm(4) driver, that limits these chips to 256 descriptors. In collaboration with: yongari
2005-11-17 10:13:18 +00:00
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
hw->mac.autoneg = DO_AUTO_NEG;
hw->phy.autoneg_wait_to_complete = FALSE;
hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type < em_mac_min) {
e1000_init_script_state_82541(&adapter->hw, TRUE);
e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Copper options */
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
if (hw->phy.media_type == e1000_media_type_copper) {
hw->phy.mdix = AUTO_ALL_MODES;
hw->phy.disable_polarity_correction = FALSE;
hw->phy.ms_type = EM_MASTER_SLAVE;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
/*
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
* Set the frame limits assuming
* standard ethernet sized frames.
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
*/
Push max_frame_len back into iflib so that jumbo frame sizes work. Reported by: pho
2017-02-17 21:30:16 +00:00
scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
Refresh on the shared code for the E1000 drivers. - bear with me, there are lots of white space changes, I would not do them, but I am a mere consumer of this stuff and if these drivers are to stay in shape they need to be taken. em driver changes: support for the new i217/i218 interfaces igb driver changes: - TX mq start has a quick turnaround to the stack - Link/media handling improvement - When link status changes happen the current flow control state will now be displayed. - A few white space/style changes. lem driver changes: - the shared code uncovered a bogus write to the RLPML register (which does not exist in this hardware) in the vlan code,this is removed.
2013-02-21 00:25:45 +00:00
ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
/*
* This controls when hardware reports transmit completion
* status.
Added support for 82541 and 82547 based adapters. - These have Intel gigabit PHY - 82547 uses CSA interface MFC after: 1 week
2003-03-21 21:47:31 +00:00
*/
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
hw->mac.report_tx_early = 1;
Added support for 82541 and 82547 based adapters. - These have Intel gigabit PHY - 82547 uses CSA interface MFC after: 1 week
2003-03-21 21:47:31 +00:00
Do not allocate multicast array memory in multicast filter configuration function. For failed memory allocations, em(4)/lem(4) called panic(9) which is not acceptable on production box. igb(4)/ixgb(4)/ix(4) allocated the required memory in stack which consumed 768 bytes of stack memory which looks too big. To address these issues, allocate multicast array memory in device attach time and make multicast configuration success under any conditions. This change also removes the excessive use of memory in stack. Reviewed by: jfv
2010-08-28 00:34:22 +00:00
/* Allocate multicast array memory. */
adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
if (adapter->mta == NULL) {
device_printf(dev, "Can not allocate multicast setup array\n");
error = ENOMEM;
goto err_late;
}
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
/* Check SOL/IDER usage */
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
if (e1000_check_reset_block(hw))
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
device_printf(dev, "PHY reset is blocked"
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
" due to SOL/IDER session.\n");
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
/* Sysctl for setting Energy Efficient Ethernet */
Change the interface to the Energy Efficient Ethernet (EEE) setting in the igb and em driver. This was necessitated by a shared code change that I was given late in the game, a data type changed from bool to int, in the last update I dealt with it by a cast, but it was pointed out (thanks jhb) that there was a potential problem with this. John suggested this safer approach, and it is fine with me... MFC after:2 days (to catch the 9.1 update)
2012-07-07 20:21:05 +00:00
hw->dev_spec.ich8lan.eee_disable = eee_setting;
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
adapter, 0, em_sysctl_eee, "I",
"Disable Energy Efficient Ethernet");
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
This delta syncs the em and igb drivers with Intel, adds header split and SCTP support into the igb driver. Various small improvements and fixes. MFC after: 2 weeks
2009-04-10 00:05:46 +00:00
/*
** Start from a known state, this is
** important in reading the nvm and
** mac from that.
*/
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
e1000_reset_hw(hw);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Make sure we have a good EEPROM before we read from it */
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
if (e1000_validate_nvm_checksum(hw) < 0) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/*
** Some PCI-E parts fail the first check due to
** the link being in sleep state, call it again,
** if it fails a second time its a real issue.
*/
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
if (e1000_validate_nvm_checksum(hw) < 0) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
device_printf(dev,
"The EEPROM Checksum Is Not Valid\n");
error = EIO;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
goto err_late;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
}
}
- Use IO mode to reset the controller (82544 and beyond) - Read the Mac address only once during attach. (This fixes the failover issue observed using the bonding driver) MFC after: 3 days
2002-07-16 16:55:03 +00:00
/* Copy the permanent MAC address out of the EEPROM */
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
if (e1000_read_mac_addr(hw) < 0) {
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
device_printf(dev, "EEPROM read error while reading MAC"
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
" address\n");
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
error = EIO;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
goto err_late;
- Use IO mode to reset the controller (82544 and beyond) - Read the Mac address only once during attach. (This fixes the failover issue observed using the bonding driver) MFC after: 3 days
2002-07-16 16:55:03 +00:00
}
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
if (!em_is_valid_ether_addr(hw->mac.addr)) {
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
device_printf(dev, "Invalid MAC address\n");
error = EIO;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
goto err_late;
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
}
Sync the E1000 shared code with Intel internal, this adds fixes, and more importantly, new I218 adapter support to the em driver. MFC after: 1 week
2014-06-26 21:33:32 +00:00
/* Disable ULP support */
e1000_disable_ulp_lpt_lp(hw, TRUE);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/*
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
* Get Wake-on-Lan and Management info for later use
*/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_get_wakeup(ctx);
iflib_set_mac(ctx, hw->mac.addr);
return (0);
err_late:
em_release_hw_control(adapter);
err_pci:
em_free_pci_resources(ctx);
free(adapter->mta, M_DEVBUF);
return (error);
}
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int
em_if_attach_post(if_ctx_t ctx)
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct e1000_hw *hw = &adapter->hw;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
int error = 0;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Setup OS specific network interface */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
error = em_setup_interface(ctx);
if (error != 0) {
Do not call voluntary panic(9) in case of if_alloc() failure. Reviewed by: jfv
2010-08-28 00:09:19 +00:00
goto err_late;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_reset(ctx);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Initialize statistics */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_update_stats_counters(adapter);
This delta updates the em driver to version 7.2.2 which has been undergoing test for some weeks. This improves the RX mbuf handling to avoid system hang due to depletion. Thanks to all those who have been testing the code, and to Beezar Liu for the design changes. Next the igb driver is updated for similar RX changes, but also to add new features support for our upcoming i350 family of adapters. MFC after a week
2011-03-18 18:54:00 +00:00
hw->mac.get_link_status = 1;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_update_admin_status(ctx);
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
em_add_hw_stats(adapter);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
/* Non-AMT based hardware can now take control from firmware */
if (adapter->has_manage && !adapter->has_amt)
em_get_hw_control(adapter);
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
INIT_DEBUGOUT("em_if_attach_post: end");
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
return (error);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
err_late:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
em_release_hw_control(adapter);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_free_pci_resources(ctx);
em_if_queues_free(ctx);
Do not allocate multicast array memory in multicast filter configuration function. For failed memory allocations, em(4)/lem(4) called panic(9) which is not acceptable on production box. igb(4)/ixgb(4)/ix(4) allocated the required memory in stack which consumed 768 bytes of stack memory which looks too big. To address these issues, allocate multicast array memory in device attach time and make multicast configuration success under any conditions. This change also removes the excessive use of memory in stack. Reviewed by: jfv
2010-08-28 00:34:22 +00:00
free(adapter->mta, M_DEVBUF);
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (error);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
/*********************************************************************
* Device removal routine
*
* The detach entry point is called when the driver is being removed.
* This routine stops the adapter and deallocates all the resources
* that were allocated for driver operation.
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
*
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
* return 0 on success, positive on failure
*********************************************************************/
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_detach(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
INIT_DEBUGOUT("em_detach: begin");
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
e1000_phy_hw_reset(&adapter->hw);
em_release_manageability(adapter);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
em_release_hw_control(adapter);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_free_pci_resources(ctx);
locking fixups: o correct recursive locking when polling and in em_82547_move_tail o destroy mutex on detach o add EM_LOCK_ASSERT and similar macros for creating+deleteing the mtx Submitted by: Daniel Eischen <eischen@vigrid.com>
2003-10-10 23:14:21 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (0);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
/*********************************************************************
*
* Shutdown entry point
- Move to array based indexing for TX/RX descriptor/buffer management - Added support for ITR (interrupt throttle register). This feature is available on adapters based on 82545 and above - Fixed problem with vlan support when traffic has priority bits set. (kern/45907) PR: kern/45907 MFC after: 1 week
2002-12-23 19:11:23 +00:00
*
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
**********************************************************************/
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_shutdown(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
return em_if_suspend(ctx);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
Add simple suspend and resume methods. We call em_stop() on suspend and em_init() on resume. With this change the network is ready right after resume, without half minute lag. Tested by: Jacques Garrigue
2005-12-26 10:39:21 +00:00
/*
* Suspend/resume device methods.
*/
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_suspend(if_ctx_t ctx)
Add simple suspend and resume methods. We call em_stop() on suspend and em_init() on resume. With this change the network is ready right after resume, without half minute lag. Tested by: Jacques Garrigue
2005-12-26 10:39:21 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Fix for a reported panic in certain circumstances. When calling em_stop() now make sure the TX lock is held as well as CORE.
2007-11-26 19:47:03 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_release_manageability(adapter);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
em_release_hw_control(adapter);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_enable_wakeup(ctx);
return (0);
Add simple suspend and resume methods. We call em_stop() on suspend and em_init() on resume. With this change the network is ready right after resume, without half minute lag. Tested by: Jacques Garrigue
2005-12-26 10:39:21 +00:00
}
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_resume(if_ctx_t ctx)
Add simple suspend and resume methods. We call em_stop() on suspend and em_init() on resume. With this change the network is ready right after resume, without half minute lag. Tested by: Jacques Garrigue
2005-12-26 10:39:21 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Add simple suspend and resume methods. We call em_stop() on suspend and em_init() on resume. With this change the network is ready right after resume, without half minute lag. Tested by: Jacques Garrigue
2005-12-26 10:39:21 +00:00
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
if (adapter->hw.mac.type == e1000_pch2lan)
e1000_resume_workarounds_pchlan(&adapter->hw);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_init(ctx);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
em_init_manageability(adapter);
Fix a few issues with transmit handling in em(4) and igb(4): - Do not define the foo_start() methods or set if_start in the ifnet if multiq transmit is enabled. Also, set if_transmit and if_qflush before ether_ifattach rather than after when multiq transmit is enabled. This helps to ensure that the drivers never try to mix different transmit methods. - Properly restart transmit during resume. igb(4) was not restarting it at all, and em(4) was restarting even if the link was down and was calling the wrong method if multiq transmit was enabled. - Remove all the 'more' handling for transmit completions. Transmit completion processing does not have a processing limit, so it always runs to completion and never has more work to do when it returns. Instead, the previous code was returning 'true' anytime there were packets in the queue that weren't still in the process of being transmitted. The effect was that the driver would continuously reschedule a task to process TX completions in effect running at 100% CPU polling the hardware until it finished transmitting all of the packets in the ring. Now it will just wait for the next TX completion interrupt. - Restart packet transmission when the link becomes active. - Fix the MSI-X queue interrupt handlers to restart packet transmission if there are pending packets in the relevant software queue (IFQ or buf_ring) after processing TX completions. This is the root cause for the OACTIVE hangs as if the MSI-X queue handler drained all the pending packets from the TX ring, nothing would ever restart it. As such, remove some previously-added workarounds to reschedule a task to poll the TX ring anytime OACTIVE was set. Tested by: sbruno Reviewed by: jfv MFC after: 1 week
2012-03-30 19:54:48 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return(0);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
Add simple suspend and resume methods. We call em_stop() on suspend and em_init() on resume. With this change the network is ready right after resume, without half minute lag. Tested by: Jacques Garrigue
2005-12-26 10:39:21 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int
em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
int max_frame_size;
struct adapter *adapter = iflib_get_softc(ctx);
if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
switch (adapter->hw.mac.type) {
case e1000_82571:
case e1000_82572:
case e1000_ich9lan:
case e1000_ich10lan:
case e1000_pch2lan:
case e1000_pch_lpt:
case e1000_pch_spt:
case e1000_82574:
case e1000_82583:
case e1000_80003es2lan:
/* 9K Jumbo Frame size */
max_frame_size = 9234;
break;
case e1000_pchlan:
max_frame_size = 4096;
break;
case e1000_82542:
case e1000_ich8lan:
/* Adapters that do not support jumbo frames */
max_frame_size = ETHER_MAX_LEN;
break;
default:
if (adapter->hw.mac.type >= igb_mac_min)
max_frame_size = 9234;
else /* lem */
max_frame_size = MAX_JUMBO_FRAME_SIZE;
}
if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
return (EINVAL);
}
scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
Actually set the MTU to the requested value and fixup handling of jumbo frames. Submitted by: Matt Macy <mmacy@nextbsd.org> Reported by: pho Sponsored by: Limelight Networks
2017-03-15 14:44:59 +00:00
mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (0);
Add simple suspend and resume methods. We call em_stop() on suspend and em_init() on resume. With this change the network is ready right after resume, without half minute lag. Tested by: Jacques Garrigue
2005-12-26 10:39:21 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*********************************************************************
* Init entry point
*
* This routine is used in two ways. It is used by the stack as
* init entry point in network interface structure. It is also used
* by the driver as a hw/sw initialization routine to get to a
* consistent state.
*
* return 0 on success, positive on failure
**********************************************************************/
- Move to array based indexing for TX/RX descriptor/buffer management - Added support for ITR (interrupt throttle register). This feature is available on adapters based on 82545 and above - Fixed problem with vlan support when traffic has priority bits set. (kern/45907) PR: kern/45907 MFC after: 1 week
2002-12-23 19:11:23 +00:00
- define em_transmit and em_qflush - make buF_ring usage conditional but enabled by default Reviewed by: jfv
2009-04-14 03:36:59 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_init(if_ctx_t ctx)
- define em_transmit and em_qflush - make buF_ring usage conditional but enabled by default Reviewed by: jfv
2009-04-14 03:36:59 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
struct ifnet *ifp = iflib_get_ifp(ctx);
struct em_tx_queue *tx_que;
int i;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
INIT_DEBUGOUT("em_if_init: begin");
- define em_transmit and em_qflush - make buF_ring usage conditional but enabled by default Reviewed by: jfv
2009-04-14 03:36:59 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Get the latest mac address, User can use a LAA */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
ETHER_ADDR_LEN);
Simplify hang detection by stealing the techniques used in ixl(4) and applying them to em(4). Rely on iterations through the local timer, and the tx queue state to determine if an actual hang has occurred. Any time a descriptor is used (packet sent), the tx queue is flagged as busy. Then when txeof runs, it either clears the flag when all is clean, or resets it to 1 if ANY are cleaned, if nothing is cleaned it increments the flag. Local timer simply checks to see if busy ever reaches MAX (10, which is compile time configurable), and then sets it as HUNG, at that point there is one more timer cycle in which to have any cleans, if not a watchdog reset will occur. Differential Revision: https://reviews.freebsd.org/D2019 Submitted by: jfv Reviewed by: hiren Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks
2015-06-02 18:28:41 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Put the address into the Receive Address Array */
e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*
* With the 82571 adapter, RAR[0] may be overwritten
* when the other port is reset, we make a duplicate
* in RAR[14] for that eventuality, this assures
* the interface continues to function.
*/
if (adapter->hw.mac.type == e1000_82571) {
e1000_set_laa_state_82571(&adapter->hw, TRUE);
e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
E1000_RAR_ENTRIES - 1);
- define em_transmit and em_qflush - make buF_ring usage conditional but enabled by default Reviewed by: jfv
2009-04-14 03:36:59 +00:00
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Initialize the hardware */
em_reset(ctx);
em_if_update_admin_status(ctx);
- define em_transmit and em_qflush - make buF_ring usage conditional but enabled by default Reviewed by: jfv
2009-04-14 03:36:59 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
struct tx_ring *txr = &tx_que->txr;
txr->tx_rs_cidx = txr->tx_rs_pidx = txr->tx_cidx_processed = 0;
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Setup VLAN support, basic and offload if available */
E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Clear bad data from Rx FIFOs */
if (adapter->hw.mac.type >= igb_mac_min)
e1000_rx_fifo_flush_82575(&adapter->hw);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Configure for OS presence */
em_init_manageability(adapter);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Prepare transmit descriptors and buffers */
em_initialize_transmit_unit(ctx);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Setup Multicast table */
em_if_multi_set(ctx);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* Figure out the desired mbuf
* pool for doing jumbos
*/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.max_frame_size <= 2048)
adapter->rx_mbuf_sz = MCLBYTES;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
#ifndef CONTIGMALLOC_WORKS
else
adapter->rx_mbuf_sz = MJUMPAGESIZE;
#else
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
else if (adapter->hw.mac.max_frame_size <= 4096)
adapter->rx_mbuf_sz = MJUMPAGESIZE;
else
adapter->rx_mbuf_sz = MJUM9BYTES;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
#endif
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_initialize_receive_unit(ctx);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
/* Use real VLAN Filter support? */
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
/* Use real VLAN Filter support */
em_setup_vlan_hw_support(adapter);
else {
u32 ctrl;
ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
ctrl |= E1000_CTRL_VME;
E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
}
}
Commit the results of the typo hunt by Darren Pilgrim. This change affects documentation and comments only, no real code involved. PR: misc/101245 Submitted by: Darren Pilgrim <darren pilgrim bitfreak org> Tested by: md5(1) MFC after: 1 week
2006-08-04 07:56:35 +00:00
/* Don't lose promiscuous settings */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_set_promisc(ctx, IFF_PROMISC);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
e1000_clear_hw_cntrs_base_generic(&adapter->hw);
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
/* MSI/X configuration for 82574 */
if (adapter->hw.mac.type == e1000_82574) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
tmp |= E1000_CTRL_EXT_PBA_CLR;
E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Set the IVAR - interrupt vector routing. */
E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
igb_configure_queues(adapter);
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* this clears any pending interrupts */
E1000_READ_REG(&adapter->hw, E1000_ICR);
E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
/* AMT based hardware can now take control from firmware */
if (adapter->has_manage && adapter->has_amt)
em_get_hw_control(adapter);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Set Energy Efficient Ethernet */
if (adapter->hw.mac.type >= igb_mac_min &&
adapter->hw.phy.media_type == e1000_media_type_copper) {
if (adapter->hw.mac.type == e1000_i354)
e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
else
e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
}
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
}
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*********************************************************************
*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* Fast Legacy/MSI Combined Interrupt Service routine
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
*
*********************************************************************/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
int
em_intr(void *arg)
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = arg;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if_ctx_t ctx = adapter->ctx;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
u32 reg_icr;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->intr_type != IFLIB_INTR_LEGACY)
goto skip_stray;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* Hot eject? */
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (reg_icr == 0xffffffff)
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
return FILTER_STRAY;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* Definitely not our interrupt. */
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (reg_icr == 0x0)
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
return FILTER_STRAY;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*
* Starting with the 82571 chip, bit 31 should be used to
* determine whether the interrupt belongs to us.
*/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if (adapter->hw.mac.type >= e1000_82571 &&
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
(reg_icr & E1000_ICR_INT_ASSERTED) == 0)
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
return FILTER_STRAY;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
skip_stray:
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/* Link status change */
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
adapter->hw.mac.get_link_status = 1;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
iflib_admin_intr_deferred(ctx);
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (reg_icr & E1000_ICR_RXO)
adapter->rx_overruns++;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (FILTER_SCHEDULE_THREAD);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
}
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
static void
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
igb_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
}
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
static void
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
}
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static void
igb_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
{
E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
}
static void
em_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
{
E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (adapter->hw.mac.type >= igb_mac_min)
igb_rx_enable_queue(adapter, rxq);
else
em_rx_enable_queue(adapter, rxq);
return (0);
}
static int
em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
{
struct adapter *adapter = iflib_get_softc(ctx);
struct em_tx_queue *txq = &adapter->tx_queues[txqid];
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type >= igb_mac_min)
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
igb_tx_enable_queue(adapter, txq);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
else
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_tx_enable_queue(adapter, txq);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
return (0);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
/*********************************************************************
*
* MSIX RX Interrupt Service routine
*
**********************************************************************/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int
em_msix_que(void *arg)
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct em_rx_queue *que = arg;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
++que->irqs;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (FILTER_SCHEDULE_THREAD);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
}
/*********************************************************************
*
* MSIX Link Fast Interrupt Service routine
*
**********************************************************************/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
em_msix_link(void *arg)
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = arg;
u32 reg_icr;
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
++adapter->link_irq;
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
MPASS(adapter->hw.back != NULL);
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
Bump rx_overruns when indicated by the ICR mask. PR: 199716 MFC after: 3 days Sponsored by: Limelight Networks
2015-05-22 17:01:43 +00:00
if (reg_icr & E1000_ICR_RXO)
adapter->rx_overruns++;
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
em_handle_link(adapter->ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
} else {
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_IMS,
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
EM_MSIX_LINK | E1000_IMS_LSC);
if (adapter->hw.mac.type >= igb_mac_min)
E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
/*
* Because we must read the ICR for this interrupt
* it may clear other causes using autoclear, for
* this reason we simply create a soft interrupt
* for all these vectors.
*/
if (reg_icr && adapter->hw.mac.type < igb_mac_min) {
E1000_WRITE_REG(&adapter->hw,
E1000_ICS, adapter->ims);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
}
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (FILTER_HANDLED);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_handle_link(void *context)
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if_ctx_t ctx = context;
struct adapter *adapter = iflib_get_softc(ctx);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
adapter->hw.mac.get_link_status = 1;
iflib_admin_intr_deferred(ctx);
This delta syncs the em and igb drivers with Intel, adds header split and SCTP support into the igb driver. Various small improvements and fixes. MFC after: 2 weeks
2009-04-10 00:05:46 +00:00
}
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
/*********************************************************************
*
* Media Ioctl callback
*
* This routine is called whenever the user queries the status of
* the interface using ifconfig.
*
**********************************************************************/
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
u_char fiber_type = IFM_1000_SX;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
INIT_DEBUGOUT("em_if_media_status: begin");
iflib_admin_intr_deferred(ctx);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
ifmr->ifm_status = IFM_AVALID;
ifmr->ifm_active = IFM_ETHER;
Add locking to the ifmedia callouts. Reviewed by: jvogel, yongari
2006-09-01 16:08:36 +00:00
if (!adapter->link_active) {
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
return;
Add locking to the ifmedia callouts. Reviewed by: jvogel, yongari
2006-09-01 16:08:36 +00:00
}
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
ifmr->ifm_status |= IFM_ACTIVE;
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
(adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type == e1000_82545)
fiber_type = IFM_1000_LX;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
ifmr->ifm_active |= fiber_type | IFM_FDX;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
} else {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
switch (adapter->link_speed) {
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
case 10:
ifmr->ifm_active |= IFM_10_T;
break;
case 100:
ifmr->ifm_active |= IFM_100_TX;
break;
case 1000:
ifmr->ifm_active |= IFM_1000_T;
break;
}
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
if (adapter->link_duplex == FULL_DUPLEX)
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
ifmr->ifm_active |= IFM_FDX;
else
ifmr->ifm_active |= IFM_HDX;
}
}
/*********************************************************************
*
* Media Ioctl callback
*
* This routine is called when the user changes speed/duplex using
* media/mediopt option with ifconfig.
*
**********************************************************************/
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_media_change(if_ctx_t ctx)
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct ifmedia *ifm = iflib_get_media(ctx);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
INIT_DEBUGOUT("em_if_media_change: begin");
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (EINVAL);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
switch (IFM_SUBTYPE(ifm->ifm_media)) {
case IFM_AUTO:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.mac.autoneg = DO_AUTO_NEG;
adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
break;
Merge in new driver from Intel, version 6.1.4. It adds support for 82571EB quad port copper NIC and has few minor fixes. Details: - if_em.c. Merged manually, viewing diff between new vendor driver and previous one. - if_em_hw.c. Dropped in from vendor, and then restored revision 1.15.
2006-08-11 10:58:24 +00:00
case IFM_1000_LX:
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
case IFM_1000_SX:
case IFM_1000_T:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.mac.autoneg = DO_AUTO_NEG;
adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
break;
case IFM_100_TX:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.mac.autoneg = FALSE;
adapter->hw.phy.autoneg_advertised = 0;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
else
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
break;
case IFM_10_T:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.mac.autoneg = FALSE;
adapter->hw.phy.autoneg_advertised = 0;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
else
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
break;
default:
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
device_printf(adapter->dev, "Unsupported media type\n");
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_init(ctx);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (0);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_set_promisc(if_ctx_t ctx, int flags)
- Bus DMA'fy the driver - Use htole* macros where appropriate so that the driver could work on non-x86 architectures - Use m_getcl() instead of MGETHDR/MCLGET macros Submitted by: sam (Sam Leffler)
2003-05-02 21:17:08 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
u32 reg_rctl;
Correct a bug in the if_em driver relating to the use of vlans with promiscuous mode introduced in 1.45, which programs the em card not to strip or prepend tags when in promiscuous mode without also modifying behavior to manually prepend a vlan header in the event that the card isn't doing it on transmit. Due to a feature of card operation, if the global VLAN prepend/strip register isn't set, setting the VLAN tag flag on individual packet descriptors will cause the packet to be transmitted using ISL encapsulation rather than 802.1Q VLAN encapsulation. This fix causes em_encap() to prepend the header by tracking whether the card is configured to temporarily disable prepending/stripping due to promiscuous mode. As a result, entering promiscuous mode on the parent em interface no longer causes vlans to appear to "wedge" or transmit ISL-encapsulated frames, which typically will not be configured/spoken by the other endpoints on the VLAN trunk. This bug may also exist in other drivers, and the additional vlan encapsulation logic should be abstracted and centralized in if_vlan.c if so. RELENG_5_3 candidate. MFC after: 1 week Tested by: pjd, rwatson Reported by: astesin at ukrtelecom dot net Reported by: Mike Tancsa <mike at sentex dot net> Reported by: Iasen Kostov <tbyte at OTEL dot net>
2004-11-12 11:03:07 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_disable_promisc(ctx);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (flags & IFF_PROMISC) {
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
/* Turn this on if you want to see bad packets */
if (em_debug_sbp)
reg_rctl |= E1000_RCTL_SBP;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
} else if (flags & IFF_ALLMULTI) {
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
reg_rctl |= E1000_RCTL_MPE;
reg_rctl &= ~E1000_RCTL_UPE;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
Implement new ETHER_BPF_MTAP macro. Roll back the various changes made to accommodate the chip being in promiscuous mode while offloading VLAN tag processing to the hardware. We can now properly handle the absence of VLAN tags from hardware stripping. Reviewed by: rwatson, andre MFC after: 1 month
2006-11-18 23:18:43 +00:00
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (0);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_disable_promisc(if_ctx_t ctx)
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
struct ifnet *ifp = iflib_get_ifp(ctx);
u32 reg_rctl;
int mcnt = 0;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
reg_rctl &= (~E1000_RCTL_UPE);
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if (if_getflags(ifp) & IFF_ALLMULTI)
Correct the multicast handling in the E1000 drivers as was done in ixgbe, thanks to Mike Karels for this fix. When exiting promiscuous mode MPE bit was being unconditionally cleared, this should not be done if we are in MAX multicast groups.
2013-04-03 23:39:54 +00:00
mcnt = MAX_NUM_MULTICAST_ADDRESSES;
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
else
mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
Correct the multicast handling in the E1000 drivers as was done in ixgbe, thanks to Mike Karels for this fix. When exiting promiscuous mode MPE bit was being unconditionally cleared, this should not be done if we are in MAX multicast groups.
2013-04-03 23:39:54 +00:00
/* Don't disable if in MAX groups */
if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
reg_rctl &= (~E1000_RCTL_MPE);
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
reg_rctl &= (~E1000_RCTL_SBP);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
/*********************************************************************
* Multicast Update
*
* This routine is called whenever multicast address list is updated.
*
**********************************************************************/
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_multi_set(if_ctx_t ctx)
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
struct ifnet *ifp = iflib_get_ifp(ctx);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
u32 reg_rctl = 0;
This delta syncs the em and igb drivers with Intel, adds header split and SCTP support into the igb driver. Various small improvements and fixes. MFC after: 2 weeks
2009-04-10 00:05:46 +00:00
u8 *mta; /* Multicast array memory */
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
int mcnt = 0;
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
IOCTL_DEBUGOUT("em_set_multi: begin");
Do not allocate multicast array memory in multicast filter configuration function. For failed memory allocations, em(4)/lem(4) called panic(9) which is not acceptable on production box. igb(4)/ixgb(4)/ix(4) allocated the required memory in stack which consumed 768 bytes of stack memory which looks too big. To address these issues, allocate multicast array memory in device attach time and make multicast configuration success under any conditions. This change also removes the excessive use of memory in stack. Reviewed by: jfv
2010-08-28 00:34:22 +00:00
mta = adapter->mta;
bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (adapter->hw.mac.type == e1000_82542 &&
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.revision_id == E1000_REVISION_2) {
reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
e1000_pci_clear_mwi(&adapter->hw);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
reg_rctl |= E1000_RCTL_RST;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
msec_delay(5);
}
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
reg_rctl |= E1000_RCTL_MPE;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
} else
This delta syncs the em and igb drivers with Intel, adds header split and SCTP support into the igb driver. Various small improvements and fixes. MFC after: 2 weeks
2009-04-10 00:05:46 +00:00
e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
- Move to array based indexing for TX/RX descriptor/buffer management - Added support for ITR (interrupt throttle register). This feature is available on adapters based on 82545 and above - Fixed problem with vlan support when traffic has priority bits set. (kern/45907) PR: kern/45907 MFC after: 1 week
2002-12-23 19:11:23 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (adapter->hw.mac.type == e1000_82542 &&
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.revision_id == E1000_REVISION_2) {
reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
reg_rctl &= ~E1000_RCTL_RST;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
msec_delay(5);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
e1000_pci_set_mwi(&adapter->hw);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
}
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
/*********************************************************************
* Timer routine
*
* This routine checks for link status and updates statistics.
*
**********************************************************************/
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_timer(if_ctx_t ctx, uint16_t qid)
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct em_rx_queue *que;
int i;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
int trigger = 0;
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (qid != 0)
Only trigger em_local_timer on queue index 0. This was causing continuous em_local_timer() executions during normal operation and was very likely to cause a lock up on igb(4) devices. Submitted by: Matt Macy (mmacy@nextbsd.org) Reported by: jtl Reviewed by: gallatin Sponsored by: Limelight Networks & Netflix
2017-02-12 23:06:41 +00:00
return;
Do not update stats counter in SWI context. Defer to the already existing admin thread. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-21 17:42:54 +00:00
iflib_admin_intr_deferred(ctx);
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
/* Reset LAA into RAR[0] on 82571 */
if ((adapter->hw.mac.type == e1000_82571) &&
e1000_get_laa_state_82571(&adapter->hw))
e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
if (adapter->hw.mac.type < em_mac_min)
lem_smartspeed(adapter);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
Change the refresh_mbuf logic slightly, add an inline to calculate the outstanding descriptors that need to be refreshed at any time, and use THAT in rxeof to determine if refreshing needs to be done. Also change the local_timer to simply fire off the appropriate interrupt rather than schedule a tasklet, its simpler. MFC in two weeks
2011-04-01 18:48:31 +00:00
/* Mask to use in the irq trigger */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->intr_type == IFLIB_INTR_MSIX) {
for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++)
trigger |= que->eims;
} else {
Change the refresh_mbuf logic slightly, add an inline to calculate the outstanding descriptors that need to be refreshed at any time, and use THAT in rxeof to determine if refreshing needs to be done. Also change the local_timer to simply fire off the appropriate interrupt rather than schedule a tasklet, its simpler. MFC in two weeks
2011-04-01 18:48:31 +00:00
trigger = E1000_ICS_RXDMT0;
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
}
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_update_admin_status(if_ctx_t ctx)
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
struct e1000_hw *hw = &adapter->hw;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct ifnet *ifp = iflib_get_ifp(ctx);
device_t dev = iflib_get_dev(ctx);
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
u32 link_check, thstat, ctrl;
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
link_check = thstat = ctrl = 0;
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
/* Get the cached link value or read phy for real */
switch (hw->phy.media_type) {
case e1000_media_type_copper:
if (hw->mac.get_link_status) {
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
if (hw->mac.type == e1000_pch_spt)
msec_delay(50);
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
/* Do the work to read phy */
e1000_check_for_link(hw);
link_check = !hw->mac.get_link_status;
This delta is primarily a fix for es2lan devices that will sometimes fail to initialize problem due to a lock contention with management hardware. However, in order to deliver that fix it was necessary to take a shared code update as a whole, and this required scattered changes in the core code to be compatible. The em driver now has VLAN HW support added as the igb driver had previously. MFC after: ASAP - in time for 7.1 RELEASE
2008-11-26 23:57:23 +00:00
if (link_check) /* ESB2 fix */
e1000_cfg_on_link_up(hw);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
} else {
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
link_check = TRUE;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
break;
case e1000_media_type_fiber:
e1000_check_for_link(hw);
link_check = (E1000_READ_REG(hw, E1000_STATUS) &
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
E1000_STATUS_LU);
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
break;
case e1000_media_type_internal_serdes:
e1000_check_for_link(hw);
link_check = adapter->hw.mac.serdes_has_link;
break;
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
/* VF device is type_unknown */
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
case e1000_media_type_unknown:
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
e1000_check_for_link(hw);
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
link_check = !hw->mac.get_link_status;
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
/* FALLTHROUGH */
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
default:
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
break;
}
Call if_link_state_change() when link status changes. PR: kern/76890 Reviewed by: rwatson, sam
2005-02-04 18:36:04 +00:00
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
/* Check for thermal downshift or shutdown */
if (hw->mac.type == e1000_i350) {
thstat = E1000_READ_REG(hw, E1000_THSTAT);
ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
}
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
/* Now check for a transition */
if (link_check && (adapter->link_active == 0)) {
e1000_get_speed_and_duplex(hw, &adapter->link_speed,
&adapter->link_duplex);
/* Check if we must disable SPEED_MODE bit on PCI-E */
if ((adapter->link_speed != SPEED_1000) &&
((hw->mac.type == e1000_82571) ||
(hw->mac.type == e1000_82572))) {
int tarc0;
tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
tarc0 &= ~TARC_SPEED_MODE_BIT;
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
if (bootverbose)
device_printf(dev, "Link is up %d Mbps %s\n",
adapter->link_speed,
((adapter->link_duplex == FULL_DUPLEX) ?
"Full Duplex" : "Half Duplex"));
adapter->link_active = 1;
adapter->smartspeed = 0;
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if_setbaudrate(ifp, adapter->link_speed * 1000000);
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) &&
(thstat & E1000_THSTAT_LINK_THROTTLE))
device_printf(dev, "Link: thermal downshift\n");
/* Delay Link Up for Phy update */
if (((hw->mac.type == e1000_i210) ||
(hw->mac.type == e1000_i211)) &&
(hw->phy.id == I210_I_PHY_ID))
msec_delay(I210_LINK_DELAY);
/* Reset if the media type changed. */
if ((hw->dev_spec._82575.media_changed) &&
(adapter->hw.mac.type >= igb_mac_min)) {
hw->dev_spec._82575.media_changed = false;
adapter->flags |= IGB_MEDIA_RESET;
em_reset(ctx);
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
iflib_link_state_change(ctx, LINK_STATE_UP, ifp->if_baudrate);
printf("Link state changed to up\n");
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
} else if (!link_check && (adapter->link_active == 1)) {
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if_setbaudrate(ifp, 0);
adapter->link_speed = 0;
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
adapter->link_duplex = 0;
if (bootverbose)
device_printf(dev, "Link is Down\n");
adapter->link_active = 0;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
iflib_link_state_change(ctx, LINK_STATE_DOWN, ifp->if_baudrate);
printf("link state changed to down\n");
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
Do not update stats counter in SWI context. Defer to the already existing admin thread. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-21 17:42:54 +00:00
em_update_stats_counters(adapter);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
/*********************************************************************
*
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
* This routine disables all traffic on the adapter by issuing a
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
* global reset on the MAC and deallocates TX/RX buffers.
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*
Fix for a reported panic in certain circumstances. When calling em_stop() now make sure the TX lock is held as well as CORE.
2007-11-26 19:47:03 +00:00
* This routine should always be called with BOTH the CORE
* and TX locks.
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
**********************************************************************/
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_stop(if_ctx_t ctx)
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
Add support for new devices. Bug Fixes: - Allow users to use LAA - Remember promiscuous mode settings while bridging - Allow gratuitous arp's to be sent PR: 52966/54488 MFC after: 1 week
2003-08-27 21:52:37 +00:00
INIT_DEBUGOUT("em_stop: begin");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
e1000_reset_hw(&adapter->hw);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type >= e1000_82544)
igb(4) enable WOL features for this class of devices. PR: 208343 Submitted by: Kaho Tashikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-01-22 18:04:57 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
Hook the identification LEDs of igb(4), lem(4) and em(4) devices up with led(4) so they can be lit or f.e. made blink via `echo f2 > /dev/led/em0` for localization purposes. Approved by: jfv MFC afer: 1 week (after r205869)
2010-03-31 20:43:24 +00:00
e1000_led_off(&adapter->hw);
e1000_cleanup_led(&adapter->hw);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/*********************************************************************
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*
* Determine hardware revision.
*
**********************************************************************/
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_identify_hardware(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
device_t dev = iflib_get_dev(ctx);
struct adapter *adapter = iflib_get_softc(ctx);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Make sure our PCI config space has the necessary stuff set */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Save off the information about this board */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
adapter->hw.vendor_id = pci_get_vendor(dev);
adapter->hw.device_id = pci_get_device(dev);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
adapter->hw.subsystem_vendor_id =
pci_read_config(dev, PCIR_SUBVEND_0, 2);
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
adapter->hw.subsystem_device_id =
pci_read_config(dev, PCIR_SUBDEV_0, 2);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Do Shared Code Init and Setup */
if (e1000_set_mac_type(&adapter->hw)) {
device_printf(dev, "Setup init failure\n");
return;
}
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_allocate_pci_resources(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
device_t dev = iflib_get_dev(ctx);
int rid, val;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Make em(4) work on big-endian architectures. - disable jumbo frame support on strict alignment architectures due to the limitation of hardware. The driver needs a fix-up code for RX side. The fix will show up in near future. - fix endian issue for 82544 on PCI-X bus. I couldn't test this as I don't have the NIC/hardware. - prefer PCIR_BAR to hardcoded EM_MMBA. - Properly checks for for 64bit BAR [1] - replace inl/outl with bus_space(9) [1] - fix endian issue on VLAN handling. - reorder header files and remove unnecessary one. Reviewed by: cognet No response from: pdeuskar, tackerman Obtained from: OpenBSD [1]
2005-11-09 08:43:18 +00:00
rid = PCIR_BAR(0);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
&rid, RF_ACTIVE);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
if (adapter->memory == NULL) {
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
device_printf(dev, "Unable to allocate bus resource: memory\n");
return (ENXIO);
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->osdep.mem_bus_space_handle =
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
rman_get_bushandle(adapter->memory);
adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Only older adapters use IO mapping */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (adapter->hw.mac.type < em_mac_min &&
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
adapter->hw.mac.type > e1000_82543) {
/* Figure our where our IO BAR is ? */
for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
val = pci_read_config(dev, rid, 4);
if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
adapter->io_rid = rid;
break;
}
rid += 4;
/* check for 64bit BAR */
if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
rid += 4;
}
if (rid >= PCIR_CIS) {
device_printf(dev, "Unable to locate IO BAR\n");
return (ENXIO);
}
adapter->ioport = bus_alloc_resource_any(dev,
SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
if (adapter->ioport == NULL) {
device_printf(dev, "Unable to allocate bus resource: "
"ioport\n");
return (ENXIO);
}
adapter->hw.io_base = 0;
adapter->osdep.io_bus_space_tag =
rman_get_bustag(adapter->ioport);
adapter->osdep.io_bus_space_handle =
rman_get_bushandle(adapter->ioport);
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
adapter->hw.back = &adapter->osdep;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
return (0);
}
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
/*********************************************************************
*
* Setup the MSIX Interrupt handlers
*
**********************************************************************/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static int
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_if_msix_intr_assign(if_ctx_t ctx, int msix)
Fix the interrupt race for real. Don't register the interrupt until after the the interface has been configured. I'm not sure how this could ever have worked before, but it should be fixed now. Also break out the interrupt degresitration function into it's own step.
2006-01-13 08:18:04 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct em_rx_queue *rx_que = adapter->rx_queues;
struct em_tx_queue *tx_que = adapter->tx_queues;
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
int error, rid, i, vector = 0, rx_vectors;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
char buf[16];
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* First set up ring resources */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
rid = vector + 1;
snprintf(buf, sizeof(buf), "rxq%d", i);
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (error) {
device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
adapter->rx_num_queues = i + 1;
goto fail;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
rx_que->msix = vector;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* Set the bit to enable interrupt
* in E1000_IMS -- bits 20 and 21
* are for RX0 and RX1, note this has
* NOTHING to do with the MSIX vector
*/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type == e1000_82574) {
rx_que->eims = 1 << (20 + i);
adapter->ims |= rx_que->eims;
adapter->ivars |= (8 | rx_que->msix) << (i * 4);
} else if (adapter->hw.mac.type == e1000_82575)
rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
else
rx_que->eims = 1 << vector;
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
}
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
rx_vectors = vector;
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
vector = 0;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
rid = vector + 1;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
snprintf(buf, sizeof(buf), "txq%d", i);
tx_que = &adapter->tx_queues[i];
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
iflib_softirq_alloc_generic(ctx, rid, IFLIB_INTR_TX, tx_que, tx_que->me, buf);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
tx_que->msix = (vector % adapter->tx_num_queues);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/*
* Set the bit to enable interrupt
* in E1000_IMS -- bits 22 and 23
* are for TX0 and TX1, note this has
* NOTHING to do with the MSIX vector
*/
if (adapter->hw.mac.type == e1000_82574) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
tx_que->eims = 1 << (22 + i);
adapter->ims |= tx_que->eims;
adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
Add missing 'else' to 3-state conditional during setup of interrupts. We don't want to overwrite the 82574 interrupt setup with a different configuration. PR: 218041 Submitted by: razmyslov@viva64.com
2017-03-24 14:25:56 +00:00
} else if (adapter->hw.mac.type == e1000_82575) {
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
tx_que->eims = E1000_EICR_TX_QUEUE0 << (i % adapter->tx_num_queues);
Add missing 'else' to 3-state conditional during setup of interrupts. We don't want to overwrite the 82574 interrupt setup with a different configuration. PR: 218041 Submitted by: razmyslov@viva64.com
2017-03-24 14:25:56 +00:00
} else {
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
tx_que->eims = 1 << (i % adapter->tx_num_queues);
Add missing 'else' to 3-state conditional during setup of interrupts. We don't want to overwrite the 82574 interrupt setup with a different configuration. PR: 218041 Submitted by: razmyslov@viva64.com
2017-03-24 14:25:56 +00:00
}
This is the merge of the Intel 6.2.9 driver. It provides all new shared code, new device support, and it is hoped a more stable driver for 6.2. RELEASE. This checkin was discussed and approved today by RE, scottl, jhb, and pdeuskar
2006-10-28 00:47:55 +00:00
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Link interrupt */
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
rid = rx_vectors + 1;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
if (error) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
goto fail;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
adapter->linkvec = rx_vectors;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type < igb_mac_min) {
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
adapter->ivars |= (8 | rx_vectors) << 16;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
adapter->ivars |= 0x80000000;
}
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
return (0);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
fail:
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
iflib_irq_free(ctx, &adapter->irq);
rx_que = adapter->rx_queues;
for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
iflib_irq_free(ctx, &rx_que->que_irq);
return (error);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
}
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
igb_configure_queues(struct adapter *adapter)
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct e1000_hw *hw = &adapter->hw;
struct em_rx_queue *rx_que;
struct em_tx_queue *tx_que;
u32 tmp, ivar = 0, newitr = 0;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* First turn on RSS capability */
if (adapter->hw.mac.type != e1000_82575)
E1000_WRITE_REG(hw, E1000_GPIE,
E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
E1000_GPIE_PBA | E1000_GPIE_NSICR);
/* Turn on MSIX */
switch (adapter->hw.mac.type) {
case e1000_82580:
case e1000_i350:
case e1000_i354:
case e1000_i210:
case e1000_i211:
case e1000_vfadapt:
case e1000_vfadapt_i350:
/* RX entries */
for (int i = 0; i < adapter->rx_num_queues; i++) {
u32 index = i >> 1;
ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
rx_que = &adapter->rx_queues[i];
if (i & 1) {
ivar &= 0xFF00FFFF;
ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
} else {
ivar &= 0xFFFFFF00;
ivar |= rx_que->msix | E1000_IVAR_VALID;
}
E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
}
/* TX entries */
for (int i = 0; i < adapter->tx_num_queues; i++) {
u32 index = i >> 1;
ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
tx_que = &adapter->tx_queues[i];
if (i & 1) {
ivar &= 0x00FFFFFF;
ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
} else {
ivar &= 0xFFFF00FF;
ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
}
E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
adapter->que_mask |= tx_que->eims;
}
Couple of changes, back down on last TSO change, instead make old adapter list still capable, but only PCI-E adapters are now enabled. The user can enable older PCI-X or PCI adapters using ifconfig. Secondly, Arthur Hartwig pointed out my MSI change was not working correctly, changed to something that now does. Thanks Arthur. There was also a fundamental bug in the 82575 MSIX code, the MSIX registers had to be mapped, opps :) Rubber-stamped by: Pdeuskar
2007-05-17 00:14:03 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* And for the link interrupt */
ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
adapter->link_mask = 1 << adapter->linkvec;
E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
break;
case e1000_82576:
/* RX entries */
for (int i = 0; i < adapter->rx_num_queues; i++) {
u32 index = i & 0x7; /* Each IVAR has two entries */
ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
rx_que = &adapter->rx_queues[i];
if (i < 8) {
ivar &= 0xFFFFFF00;
ivar |= rx_que->msix | E1000_IVAR_VALID;
} else {
ivar &= 0xFF00FFFF;
ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
}
E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
adapter->que_mask |= rx_que->eims;
}
/* TX entries */
for (int i = 0; i < adapter->tx_num_queues; i++) {
u32 index = i & 0x7; /* Each IVAR has two entries */
ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
tx_que = &adapter->tx_queues[i];
if (i < 8) {
ivar &= 0xFFFF00FF;
ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
} else {
ivar &= 0x00FFFFFF;
ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
}
E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
adapter->que_mask |= tx_que->eims;
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
}
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* And for the link interrupt */
ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
adapter->link_mask = 1 << adapter->linkvec;
E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
break;
case e1000_82575:
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* enable MSI-X support*/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
tmp |= E1000_CTRL_EXT_PBA_CLR;
/* Auto-Mask interrupts upon ICR read. */
tmp |= E1000_CTRL_EXT_EIAME;
tmp |= E1000_CTRL_EXT_IRCA;
E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Queues */
for (int i = 0; i < adapter->rx_num_queues; i++) {
rx_que = &adapter->rx_queues[i];
tmp = E1000_EICR_RX_QUEUE0 << i;
tmp |= E1000_EICR_TX_QUEUE0 << i;
rx_que->eims = tmp;
E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
i, rx_que->eims);
adapter->que_mask |= rx_que->eims;
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Link */
E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
E1000_EIMS_OTHER);
adapter->link_mask |= E1000_EIMS_OTHER;
default:
break;
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
}
Add MSI support to em(4), bce(4), and mpt(4). For now, we only support devices that support a maximum of 1 message, and we use that 1 message instead of the INTx rid 0 IRQ with the same interrupt handler, etc.
2006-11-15 20:04:57 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Set the starting interrupt rate */
if (em_max_interrupt_rate > 0)
newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (hw->mac.type == e1000_82575)
newitr |= newitr << 16;
else
newitr |= E1000_EITR_CNT_IGNR;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (int i = 0; i < adapter->rx_num_queues; i++) {
rx_que = &adapter->rx_queues[i];
E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
return;
}
static void
em_free_pci_resources(if_ctx_t ctx)
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
struct em_rx_queue *que = adapter->rx_queues;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
device_t dev = iflib_get_dev(ctx);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Release all msix queue resources */
if (adapter->intr_type == IFLIB_INTR_MSIX)
iflib_irq_free(ctx, &adapter->irq);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Revert r323516 (iflib rollup) This was really too big of a commit even if everything worked, but there are multiple new issues introduced in the one huge commit, so it's not worth keeping this until it's fixed. I'll work on splitting this up into logical chunks and introduce them one at a time over the next week or two. Approved by: sbruno (mentor) Sponsored by: Limelight Networks
2017-09-16 02:41:38 +00:00
for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
iflib_irq_free(ctx, &que->que_irq);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* First release all the interrupt resources */
if (adapter->memory != NULL) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
bus_release_resource(dev, SYS_RES_MEMORY,
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
PCIR_BAR(0), adapter->memory);
adapter->memory = NULL;
}
- Use IO mode to reset the controller (82544 and beyond) - Read the Mac address only once during attach. (This fixes the failover issue observed using the bonding driver) MFC after: 3 days
2002-07-16 16:55:03 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->flash != NULL) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
bus_release_resource(dev, SYS_RES_MEMORY,
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
EM_FLASH, adapter->flash);
adapter->flash = NULL;
}
if (adapter->ioport != NULL)
bus_release_resource(dev, SYS_RES_IOPORT,
adapter->io_rid, adapter->ioport);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Setup MSI or MSI/X */
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_setup_msix(if_ctx_t ctx)
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type == e1000_82574) {
em_enable_vectors_82574(ctx);
Improve the MSIX setup code in the drivers, thanks to Marius for the changes. Make sure that pci_alloc_msix() does give us the vectors we need and fall back to MSI when it doesn't, also release any that were allocated when insufficient. MFC after: 3 days
2013-08-12 22:54:38 +00:00
}
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
return (0);
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*********************************************************************
*
* Initialize the hardware to a configuration
* as specified by the adapter structure.
*
**********************************************************************/
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
lem_smartspeed(struct adapter *adapter)
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
u16 phy_tmp;
if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
adapter->hw.mac.autoneg == 0 ||
(adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
return;
if (adapter->smartspeed == 0) {
/* If Master/Slave config fault is asserted twice,
* we assume back-to-back */
e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
return;
e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
e1000_read_phy_reg(&adapter->hw,
PHY_1000T_CTRL, &phy_tmp);
if(phy_tmp & CR_1000T_MS_ENABLE) {
phy_tmp &= ~CR_1000T_MS_ENABLE;
e1000_write_phy_reg(&adapter->hw,
PHY_1000T_CTRL, phy_tmp);
adapter->smartspeed++;
if(adapter->hw.mac.autoneg &&
!e1000_copper_link_autoneg(&adapter->hw) &&
!e1000_read_phy_reg(&adapter->hw,
PHY_CONTROL, &phy_tmp)) {
phy_tmp |= (MII_CR_AUTO_NEG_EN |
MII_CR_RESTART_AUTO_NEG);
e1000_write_phy_reg(&adapter->hw,
PHY_CONTROL, phy_tmp);
}
}
}
return;
} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
/* If still no link, perhaps using 2/3 pair cable */
e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
phy_tmp |= CR_1000T_MS_ENABLE;
e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
if(adapter->hw.mac.autoneg &&
!e1000_copper_link_autoneg(&adapter->hw) &&
!e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
phy_tmp |= (MII_CR_AUTO_NEG_EN |
MII_CR_RESTART_AUTO_NEG);
e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
}
}
/* Restart process after EM_SMARTSPEED_MAX iterations */
if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
adapter->smartspeed = 0;
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
}
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
/*********************************************************************
*
* Initialize the DMA Coalescing feature
*
**********************************************************************/
static void
igb_init_dmac(struct adapter *adapter, u32 pba)
{
device_t dev = adapter->dev;
struct e1000_hw *hw = &adapter->hw;
u32 dmac, reg = ~E1000_DMACR_DMAC_EN;
u16 hwm;
u16 max_frame_size;
if (hw->mac.type == e1000_i211)
return;
max_frame_size = adapter->shared->isc_max_frame_size;
if (hw->mac.type > e1000_82580) {
if (adapter->dmac == 0) { /* Disabling it */
E1000_WRITE_REG(hw, E1000_DMACR, reg);
return;
} else
device_printf(dev, "DMA Coalescing enabled\n");
/* Set starting threshold */
E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
hwm = 64 * pba - max_frame_size / 16;
if (hwm < 64 * (pba - 6))
hwm = 64 * (pba - 6);
reg = E1000_READ_REG(hw, E1000_FCRTC);
reg &= ~E1000_FCRTC_RTH_COAL_MASK;
reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
& E1000_FCRTC_RTH_COAL_MASK);
E1000_WRITE_REG(hw, E1000_FCRTC, reg);
dmac = pba - max_frame_size / 512;
if (dmac < pba - 10)
dmac = pba - 10;
reg = E1000_READ_REG(hw, E1000_DMACR);
reg &= ~E1000_DMACR_DMACTHR_MASK;
reg = ((dmac << E1000_DMACR_DMACTHR_SHIFT)
& E1000_DMACR_DMACTHR_MASK);
/* transition to L0x or L1 if available..*/
reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
/* Check if status is 2.5Gb backplane connection
* before configuration of watchdog timer, which is
* in msec values in 12.8usec intervals
* watchdog timer= msec values in 32usec intervals
* for non 2.5Gb connection
*/
if (hw->mac.type == e1000_i354) {
int status = E1000_READ_REG(hw, E1000_STATUS);
if ((status & E1000_STATUS_2P5_SKU) &&
(!(status & E1000_STATUS_2P5_SKU_OVER)))
reg |= ((adapter->dmac * 5) >> 6);
else
reg |= (adapter->dmac >> 5);
} else {
reg |= (adapter->dmac >> 5);
}
E1000_WRITE_REG(hw, E1000_DMACR, reg);
E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
/* Set the interval before transition */
reg = E1000_READ_REG(hw, E1000_DMCTLX);
if (hw->mac.type == e1000_i350)
reg |= IGB_DMCTLX_DCFLUSH_DIS;
/*
** in 2.5Gb connection, TTLX unit is 0.4 usec
** which is 0x4*2 = 0xA. But delay is still 4 usec
*/
if (hw->mac.type == e1000_i354) {
int status = E1000_READ_REG(hw, E1000_STATUS);
if ((status & E1000_STATUS_2P5_SKU) &&
(!(status & E1000_STATUS_2P5_SKU_OVER)))
reg |= 0xA;
else
reg |= 0x4;
} else {
reg |= 0x4;
}
E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
/* free space in tx packet buffer to wake from DMA coal */
E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
(2 * max_frame_size)) >> 6);
/* make low power state decision controlled by DMA coal */
reg = E1000_READ_REG(hw, E1000_PCIEMISC);
reg &= ~E1000_PCIEMISC_LX_DECISION;
E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
} else if (hw->mac.type == e1000_82580) {
u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
E1000_WRITE_REG(hw, E1000_PCIEMISC,
reg & ~E1000_PCIEMISC_LX_DECISION);
E1000_WRITE_REG(hw, E1000_DMACR, 0);
}
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_reset(if_ctx_t ctx)
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
device_t dev = iflib_get_dev(ctx);
struct adapter *adapter = iflib_get_softc(ctx);
struct ifnet *ifp = iflib_get_ifp(ctx);
struct e1000_hw *hw = &adapter->hw;
u16 rx_buffer_size;
u32 pba;
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
INIT_DEBUGOUT("em_reset: begin");
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
/* Let the firmware know the OS is in control */
em_get_hw_control(adapter);
Added support for 82541 and 82547 based adapters. - These have Intel gigabit PHY - 82547 uses CSA interface MFC after: 1 week
2003-03-21 21:47:31 +00:00
Merge in new driver from Intel, version 6.0.5. It adds support for 80003 NICs and NICs found on ICH8 mobos, and improves support for already known chips. Details: - if_em.c. Merged manually, viewing diff between new vendor driver and previous one. This was an easy task, because most changes between 5.1.5 and 6.0.5 are bugfixes taken from FreeBSD. - if_em_hw.h. Dropped in from vendor, and then restored revisions 1.16, 1.17, 1.18. - if_em_hw.c. Dropped in from vendor, and then restored revision 1.15. - if_em_osdep.h. Added new required macros from vendor file and add a hack against define namespace mangling in if_em_hw.h. Intel made another hack, but I prefer mine.
2006-08-03 09:20:11 +00:00
/* Set up smart power down as default off on newer adapters. */
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
hw->mac.type == e1000_82572)) {
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
u16 phy_tmp = 0;
Merge in new driver from Intel, version 6.0.5. It adds support for 80003 NICs and NICs found on ICH8 mobos, and improves support for already known chips. Details: - if_em.c. Merged manually, viewing diff between new vendor driver and previous one. This was an easy task, because most changes between 5.1.5 and 6.0.5 are bugfixes taken from FreeBSD. - if_em_hw.h. Dropped in from vendor, and then restored revisions 1.16, 1.17, 1.18. - if_em_hw.c. Dropped in from vendor, and then restored revision 1.15. - if_em_osdep.h. Added new required macros from vendor file and add a hack against define namespace mangling in if_em_hw.h. Intel made another hack, but I prefer mine.
2006-08-03 09:20:11 +00:00
/* Speed up time to link by disabling smart power down. */
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
Merge in new driver from Intel, version 6.0.5. It adds support for 80003 NICs and NICs found on ICH8 mobos, and improves support for already known chips. Details: - if_em.c. Merged manually, viewing diff between new vendor driver and previous one. This was an easy task, because most changes between 5.1.5 and 6.0.5 are bugfixes taken from FreeBSD. - if_em_hw.h. Dropped in from vendor, and then restored revisions 1.16, 1.17, 1.18. - if_em_hw.c. Dropped in from vendor, and then restored revision 1.15. - if_em_osdep.h. Added new required macros from vendor file and add a hack against define namespace mangling in if_em_hw.h. Intel made another hack, but I prefer mine.
2006-08-03 09:20:11 +00:00
phy_tmp &= ~IGP02E1000_PM_SPD;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
Merge in new driver from Intel, version 6.0.5. It adds support for 80003 NICs and NICs found on ICH8 mobos, and improves support for already known chips. Details: - if_em.c. Merged manually, viewing diff between new vendor driver and previous one. This was an easy task, because most changes between 5.1.5 and 6.0.5 are bugfixes taken from FreeBSD. - if_em_hw.h. Dropped in from vendor, and then restored revisions 1.16, 1.17, 1.18. - if_em_hw.c. Dropped in from vendor, and then restored revision 1.15. - if_em_osdep.h. Added new required macros from vendor file and add a hack against define namespace mangling in if_em_hw.h. Intel made another hack, but I prefer mine.
2006-08-03 09:20:11 +00:00
}
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
/*
* Packet Buffer Allocation (PBA)
* Writing PBA sets the receive portion of the buffer
* the remainder is used for the transmit buffer.
*/
switch (hw->mac.type) {
/* Total Packet Buffer on these is 48K */
case e1000_82571:
case e1000_82572:
case e1000_80003es2lan:
pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
break;
case e1000_82573: /* 82573: Total Packet Buffer is 32K */
pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
break;
case e1000_82574:
case e1000_82583:
pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
break;
case e1000_ich8lan:
pba = E1000_PBA_8K;
break;
case e1000_ich9lan:
case e1000_ich10lan:
/* Boost Receive side for jumbo frames */
Refresh on the shared code for the E1000 drivers. - bear with me, there are lots of white space changes, I would not do them, but I am a mere consumer of this stuff and if these drivers are to stay in shape they need to be taken. em driver changes: support for the new i217/i218 interfaces igb driver changes: - TX mq start has a quick turnaround to the stack - Link/media handling improvement - When link status changes happen the current flow control state will now be displayed. - A few white space/style changes. lem driver changes: - the shared code uncovered a bogus write to the RLPML register (which does not exist in this hardware) in the vlan code,this is removed.
2013-02-21 00:25:45 +00:00
if (adapter->hw.mac.max_frame_size > 4096)
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
pba = E1000_PBA_14K;
else
pba = E1000_PBA_10K;
break;
case e1000_pchlan:
case e1000_pch2lan:
Refresh on the shared code for the E1000 drivers. - bear with me, there are lots of white space changes, I would not do them, but I am a mere consumer of this stuff and if these drivers are to stay in shape they need to be taken. em driver changes: support for the new i217/i218 interfaces igb driver changes: - TX mq start has a quick turnaround to the stack - Link/media handling improvement - When link status changes happen the current flow control state will now be displayed. - A few white space/style changes. lem driver changes: - the shared code uncovered a bogus write to the RLPML register (which does not exist in this hardware) in the vlan code,this is removed.
2013-02-21 00:25:45 +00:00
case e1000_pch_lpt:
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
case e1000_pch_spt:
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
pba = E1000_PBA_26K;
break;
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
case e1000_82575:
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
pba = E1000_PBA_32K;
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
break;
case e1000_82576:
case e1000_vfadapt:
pba = E1000_READ_REG(hw, E1000_RXPBS);
pba &= E1000_RXPBS_SIZE_MASK_82576;
break;
case e1000_82580:
case e1000_i350:
case e1000_i354:
case e1000_vfadapt_i350:
pba = E1000_READ_REG(hw, E1000_RXPBS);
pba = e1000_rxpbs_adjust_82580(pba);
break;
case e1000_i210:
case e1000_i211:
pba = E1000_PBA_34K;
break;
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
default:
Refresh on the shared code for the E1000 drivers. - bear with me, there are lots of white space changes, I would not do them, but I am a mere consumer of this stuff and if these drivers are to stay in shape they need to be taken. em driver changes: support for the new i217/i218 interfaces igb driver changes: - TX mq start has a quick turnaround to the stack - Link/media handling improvement - When link status changes happen the current flow control state will now be displayed. - A few white space/style changes. lem driver changes: - the shared code uncovered a bogus write to the RLPML register (which does not exist in this hardware) in the vlan code,this is removed.
2013-02-21 00:25:45 +00:00
if (adapter->hw.mac.max_frame_size > 8192)
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
else
pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
}
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* Special needs in case of Jumbo frames */
if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
u32 tx_space, min_tx, min_rx;
pba = E1000_READ_REG(hw, E1000_PBA);
tx_space = pba >> 16;
pba &= 0xffff;
min_tx = (adapter->hw.mac.max_frame_size +
sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
min_tx = roundup2(min_tx, 1024);
min_tx >>= 10;
min_rx = adapter->hw.mac.max_frame_size;
min_rx = roundup2(min_rx, 1024);
min_rx >>= 10;
if (tx_space < min_tx &&
((min_tx - tx_space) < pba)) {
pba = pba - (min_tx - tx_space);
/*
* if short on rx space, rx wins
* and must trump tx adjustment
*/
if (pba < min_rx)
pba = min_rx;
}
E1000_WRITE_REG(hw, E1000_PBA, pba);
}
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
if (hw->mac.type < igb_mac_min)
E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
Merge in new driver version from Intel - 3.2.18. The most important change is support for adapters based on 82571 and 82572 chips. Tested on: 82547EI on i386 Tested on: 82540EM on sparc64
2005-11-24 01:44:49 +00:00
/*
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
* These parameters control the automatic generation (Tx) and
Merge in new driver version from Intel - 3.2.18. The most important change is support for adapters based on 82571 and 82572 chips. Tested on: 82547EI on i386 Tested on: 82540EM on sparc64
2005-11-24 01:44:49 +00:00
* response (Rx) to Ethernet PAUSE frames.
* - High water mark should allow for at least two frames to be
* received after sending an XOFF.
* - Low water mark works best when it is very near the high water mark.
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
* This allows the receiver to restart by sending XON when it has
dev/e1000,ixgbe: minor spelling fixes. No functional change. Differential Revision: https://reviews.freebsd.org/D6177
2016-05-06 22:54:56 +00:00
* drained a bit. Here we use an arbitrary value of 1500 which will
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
* restart after one full frame is pulled from the buffer. There
* could be several smaller frames in the buffer and if so they will
* not trigger the XON until their total number reduces the buffer
* by 1500.
Merge in new driver version from Intel - 3.2.18. The most important change is support for adapters based on 82571 and 82572 chips. Tested on: 82547EI on i386 Tested on: 82540EM on sparc64
2005-11-24 01:44:49 +00:00
* - The pause time is fairly large at 1000 x 512ns = 512 usec.
*/
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
rx_buffer_size = (pba & 0xffff) << 10;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
hw->fc.high_water = rx_buffer_size -
Refresh on the shared code for the E1000 drivers. - bear with me, there are lots of white space changes, I would not do them, but I am a mere consumer of this stuff and if these drivers are to stay in shape they need to be taken. em driver changes: support for the new i217/i218 interfaces igb driver changes: - TX mq start has a quick turnaround to the stack - Link/media handling improvement - When link status changes happen the current flow control state will now be displayed. - A few white space/style changes. lem driver changes: - the shared code uncovered a bogus write to the RLPML register (which does not exist in this hardware) in the vlan code,this is removed.
2013-02-21 00:25:45 +00:00
roundup2(adapter->hw.mac.max_frame_size, 1024);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
hw->fc.low_water = hw->fc.high_water - 1500;
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
if (adapter->fc) /* locally set flow control value? */
hw->fc.requested_mode = adapter->fc;
else
hw->fc.requested_mode = e1000_fc_full;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
if (hw->mac.type == e1000_80003es2lan)
hw->fc.pause_time = 0xFFFF;
Merge in new driver from Intel, version 5.1.5. Adds support for some new chips and improves support for already supported ones. Some details, important for future merges: - if_em.c merged manually, viewing diff between new vendor driver and previous one. - if_em_hw.h dropped in from vendor, and then restored revisions 1.16, 1.17, 1.18. - if_em_hw.c dropped in from vendor, and then two liner change made, that restores support for two rare chips.
2006-04-06 17:09:03 +00:00
else
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
hw->fc.pause_time = EM_FC_PAUSE_TIME;
hw->fc.send_xon = TRUE;
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
/* Device specific overrides/settings */
switch (hw->mac.type) {
case e1000_pchlan:
/* Workaround: no TX flow ctrl for PCH */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
hw->fc.requested_mode = e1000_fc_rx_pause;
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
hw->fc.pause_time = 0xFFFF; /* override */
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if (if_getmtu(ifp) > ETHERMTU) {
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
hw->fc.high_water = 0x3500;
hw->fc.low_water = 0x1500;
} else {
hw->fc.high_water = 0x5000;
hw->fc.low_water = 0x3000;
}
hw->fc.refresh_time = 0x1000;
break;
case e1000_pch2lan:
Refresh on the shared code for the E1000 drivers. - bear with me, there are lots of white space changes, I would not do them, but I am a mere consumer of this stuff and if these drivers are to stay in shape they need to be taken. em driver changes: support for the new i217/i218 interfaces igb driver changes: - TX mq start has a quick turnaround to the stack - Link/media handling improvement - When link status changes happen the current flow control state will now be displayed. - A few white space/style changes. lem driver changes: - the shared code uncovered a bogus write to the RLPML register (which does not exist in this hardware) in the vlan code,this is removed.
2013-02-21 00:25:45 +00:00
case e1000_pch_lpt:
Update em(4) to 7.6.1; update igb(4) to 2.5.3. Major changes: - Add i219/i219(2) hardware support. (Found on Skylake generation and newer chipsets.) - Further to the last Skylake support diff, this one also includes support for the Lewisburg chipset (i219(3)). - Add a workaround to an igb hardware errata. All 1G server products need to have IPv6 extension header parsing turned off. This should be listed in the specification updates for current 1G server products, e.g. for i350 it's errata #37 in this document: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/ethernet-controller-i350-spec-update.pdf - Avoton (i354) PHY errata workaround added And a bunch of minor fixes, as well as #defines for things that the current em(4)/igb(4) drivers don't implement. Differential Revision: https://reviews.freebsd.org/D3162 Reviewed by: sbruno, marius, gnn Approved by: gnn MFC after: 2 weeks Sponsored by: Intel Corporation
2016-02-05 17:14:37 +00:00
case e1000_pch_spt:
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
hw->fc.high_water = 0x5C20;
hw->fc.low_water = 0x5048;
hw->fc.pause_time = 0x0650;
hw->fc.refresh_time = 0x0400;
/* Jumbos need adjusted PBA */
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if (if_getmtu(ifp) > ETHERMTU)
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
E1000_WRITE_REG(hw, E1000_PBA, 12);
else
E1000_WRITE_REG(hw, E1000_PBA, 26);
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
break;
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
case e1000_82575:
case e1000_82576:
/* 8-byte granularity */
hw->fc.low_water = hw->fc.high_water - 8;
break;
case e1000_82580:
case e1000_i350:
case e1000_i354:
case e1000_i210:
case e1000_i211:
case e1000_vfadapt:
case e1000_vfadapt_i350:
/* 16-byte granularity */
hw->fc.low_water = hw->fc.high_water - 16;
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
break;
case e1000_ich9lan:
case e1000_ich10lan:
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if (if_getmtu(ifp) > ETHERMTU) {
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
hw->fc.high_water = 0x2800;
hw->fc.low_water = hw->fc.high_water - 8;
break;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
}
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
/* FALLTHROUGH */
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
default:
if (hw->mac.type == e1000_80003es2lan)
hw->fc.pause_time = 0xFFFF;
break;
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
}
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Issue a global reset */
e1000_reset_hw(hw);
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
if (adapter->hw.mac.type >= igb_mac_min) {
E1000_WRITE_REG(hw, E1000_WUC, 0);
} else {
E1000_WRITE_REG(hw, E1000_WUFC, 0);
em_disable_aspm(adapter);
}
if (adapter->flags & IGB_MEDIA_RESET) {
e1000_setup_init_funcs(hw, TRUE);
e1000_get_bus_info(hw);
adapter->flags &= ~IGB_MEDIA_RESET;
}
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
/* and a re-init */
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
if (e1000_init_hw(hw) < 0) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
device_printf(dev, "Hardware Initialization Failed\n");
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
return;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
if (adapter->hw.mac.type >= igb_mac_min)
igb_init_dmac(adapter, pba);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
e1000_get_phy_info(hw);
e1000_check_for_link(hw);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
#define RSSKEYLEN 10
static void
em_initialize_rss_mapping(struct adapter *adapter)
{
uint8_t rss_key[4 * RSSKEYLEN];
uint32_t reta = 0;
struct e1000_hw *hw = &adapter->hw;
int i;
/*
* Configure RSS key
*/
arc4rand(rss_key, sizeof(rss_key), 0);
for (i = 0; i < RSSKEYLEN; ++i) {
uint32_t rssrk = 0;
rssrk = EM_RSSRK_VAL(rss_key, i);
E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
}
/*
* Configure RSS redirect table in following fashion:
* (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
*/
for (i = 0; i < sizeof(reta); ++i) {
uint32_t q;
q = (i % adapter->rx_num_queues) << 7;
reta |= q << (8 * i);
}
for (i = 0; i < 32; ++i)
E1000_WRITE_REG(hw, E1000_RETA(i), reta);
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
E1000_MRQC_RSS_FIELD_IPV4_TCP |
E1000_MRQC_RSS_FIELD_IPV4 |
E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
E1000_MRQC_RSS_FIELD_IPV6_EX |
E1000_MRQC_RSS_FIELD_IPV6);
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void
igb_initialize_rss_mapping(struct adapter *adapter)
{
struct e1000_hw *hw = &adapter->hw;
int i;
int queue_id;
u32 reta;
u32 rss_key[10], mrqc, shift = 0;
/* XXX? */
if (adapter->hw.mac.type == e1000_82575)
shift = 6;
/*
* The redirection table controls which destination
* queue each bucket redirects traffic to.
* Each DWORD represents four queues, with the LSB
* being the first queue in the DWORD.
*
* This just allocates buckets to queues using round-robin
* allocation.
*
* NOTE: It Just Happens to line up with the default
* RSS allocation method.
*/
/* Warning FM follows */
reta = 0;
for (i = 0; i < 128; i++) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
#ifdef RSS
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
queue_id = rss_get_indirection_to_bucket(i);
/*
* If we have more queues than buckets, we'll
* end up mapping buckets to a subset of the
* queues.
*
* If we have more buckets than queues, we'll
* end up instead assigning multiple buckets
* to queues.
*
* Both are suboptimal, but we need to handle
* the case so we don't go out of bounds
* indexing arrays and such.
*/
queue_id = queue_id % adapter->rx_num_queues;
#else
queue_id = (i % adapter->rx_num_queues);
#endif
/* Adjust if required */
queue_id = queue_id << shift;
/*
* The low 8 bits are for hash value (n+0);
* The next 8 bits are for hash value (n+1), etc.
*/
reta = reta >> 8;
reta = reta | ( ((uint32_t) queue_id) << 24);
if ((i & 3) == 3) {
E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
reta = 0;
}
}
/* Now fill in hash table */
/*
* MRQC: Multiple Receive Queues Command
* Set queuing to RSS control, number depends on the device.
*/
mrqc = E1000_MRQC_ENABLE_RSS_8Q;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
#ifdef RSS
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* XXX ew typecasting */
rss_getkey((uint8_t *) &rss_key);
#else
arc4rand(&rss_key, sizeof(rss_key), 0);
#endif
for (i = 0; i < 10; i++)
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*
* Configure the RSS fields to hash upon.
*/
mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
E1000_MRQC_RSS_FIELD_IPV4_TCP);
mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
E1000_MRQC_RSS_FIELD_IPV6_TCP);
mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
E1000_MRQC_RSS_FIELD_IPV6_UDP);
mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
/*********************************************************************
*
* Setup networking device structure and register an interface.
*
**********************************************************************/
Do not call voluntary panic(9) in case of if_alloc() failure. Reviewed by: jfv
2010-08-28 00:09:19 +00:00
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_setup_interface(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct ifnet *ifp = iflib_get_ifp(ctx);
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if_softc_ctx_t scctx = adapter->shared;
uint64_t cap = 0;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
INIT_DEBUGOUT("em_setup_interface: begin");
Increase EM_MAX_SCATTER to 64 such that the size of em_xmit()::segs[EM_MAX_SCATTER] doesn't get overrun by things like NFS that can and do shove more than 32 segs when being used with em(4) and TSO4. Update tso handling code in em_xmit() with update from jhb@ in email thread: https://lists.freebsd.org/pipermail/freebsd-net/2014-July/039306.html set ifp->if_hw_tsomax, ifp->if_hw_tsomaxsegcount & ifp->if_hw_tsomaxsegsize to appropriate values. Define a TSO workaround "magic" number of 4 that is used to avoid an alignment issue in hardware. Change a couple of integer values that were used as booleans to actual bool types. Ensure that em_enable_intr() enables the appropriate mask of interrupts and not just a hardcoded define of values. PR: 200221 199174 195078 Differential Revision: https://reviews.freebsd.org/D3192 Reviewed by: erj jhb hiren MFC after: 2 weeks Sponsored by: Limelight Networks
2015-08-16 19:43:44 +00:00
/* TSO parameters */
Add the folowing set accessor functions for recently-added members of ifnet structure: if_gethwtsomax(), if_sethwtsomax() - if_hw_tsomax if_gethwtsomaxsegcount(), if_sethwtsomaxsegcount() - if_hw_tsomaxsegcount if_gethwtsomaxsegsize(), if_sethwtsomaxsegsize() - if_hw_tsomaxsegsize Update em and vnic drivers which had already been coverted to use accessor functions for the other ifnet structure members. Reviewed by: erj Approved by: sjg (mentor) Obtained from: Juniper Networks, Inc. Differential Revision: https://reviews.freebsd.org/D8544
2017-01-31 16:12:31 +00:00
if_sethwtsomax(ifp, IP_MAXPACKET);
Fix and clean up usage of DMA and TSO segments: - At Intel it is believed that most of their products support "only" 40 DMA segments so lower {EM,IGB}_MAX_SCATTER accordingly. Actually, 40 is more than plenty to handle full size TSO packets so it doesn't make sense to further distinguish between MAC variants that really can do 64 DMA segments. Moreover, capping at 40 DMA segments limits the stack usage of {em,igb}_xmit() that - given the rare use of more than these - previously hardly was justifiable, while still being sufficient to avoid the problems seen with em(4) and EM_MAX_SCATTER set to 32. - In igb(4), pass the actually supported TSO parameters up the stack. Previously, the defaults set in if_attach_internal() were applied, i. e. a maximum of 35 TSO segments, which made supporting more than these in the driver pointless. However, this might explain why no problems were seen with IGB_MAX_SCATTER at 64. - In em(4), take the 5 m_pullup(9) invocations performed by em_xmit() in the TSO case into account when reporting TSO parameters upwards. In the worst case, each of these calls will add another mbuf and, thus, the requirement for an additional DMA segment. So for best performance, it doesn't make sense to advertize a maximum of TSO segments that typically will require defragmentation in em_xmit(). Again, this leaves enough room to handle full size TSO packets. - Drop TSO macros from if_lem.h given that corresponding MACS don't support TSO in the first place. Reviewed by: erj, sbruno, jeffrey.e.pieper_intel.com Approved by: erj MFC after: 3 days Differential Revision: https://reviews.freebsd.org/D5238
2016-02-23 01:19:26 +00:00
/* Take m_pullup(9)'s in em_xmit() w/ TSO into acount. */
Add the folowing set accessor functions for recently-added members of ifnet structure: if_gethwtsomax(), if_sethwtsomax() - if_hw_tsomax if_gethwtsomaxsegcount(), if_sethwtsomaxsegcount() - if_hw_tsomaxsegcount if_gethwtsomaxsegsize(), if_sethwtsomaxsegsize() - if_hw_tsomaxsegsize Update em and vnic drivers which had already been coverted to use accessor functions for the other ifnet structure members. Reviewed by: erj Approved by: sjg (mentor) Obtained from: Juniper Networks, Inc. Differential Revision: https://reviews.freebsd.org/D8544
2017-01-31 16:12:31 +00:00
if_sethwtsomaxsegcount(ifp, EM_MAX_SCATTER - 5);
if_sethwtsomaxsegsize(ifp, EM_TSO_SEG_SIZE);
Increase EM_MAX_SCATTER to 64 such that the size of em_xmit()::segs[EM_MAX_SCATTER] doesn't get overrun by things like NFS that can and do shove more than 32 segs when being used with em(4) and TSO4. Update tso handling code in em_xmit() with update from jhb@ in email thread: https://lists.freebsd.org/pipermail/freebsd-net/2014-July/039306.html set ifp->if_hw_tsomax, ifp->if_hw_tsomaxsegcount & ifp->if_hw_tsomaxsegsize to appropriate values. Define a TSO workaround "magic" number of 4 that is used to avoid an alignment issue in hardware. Change a couple of integer values that were used as booleans to actual bool types. Ensure that em_enable_intr() enables the appropriate mask of interrupts and not just a hardcoded define of values. PR: 200221 199174 195078 Differential Revision: https://reviews.freebsd.org/D3192 Reviewed by: erj jhb hiren MFC after: 2 weeks Sponsored by: Limelight Networks
2015-08-16 19:43:44 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Single Queue */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (adapter->tx_num_queues == 1) {
if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
if_setsendqready(ifp);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
Stylistic changes around the previous commit: - since the number of supported capabilities is growing, set bits in if_cap* in a consistent way; - unexpand(1) leading SPACE characters.
2004-05-20 11:04:09 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
cap = IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM | IFCAP_TSO4;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
cap |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
- Move to array based indexing for TX/RX descriptor/buffer management - Added support for ITR (interrupt throttle register). This feature is available on adapters based on 82545 and above - Fixed problem with vlan support when traffic has priority bits set. (kern/45907) PR: kern/45907 MFC after: 1 week
2002-12-23 19:11:23 +00:00
Tell the upper layer(s) that we support long frames. Not doing this caused the vlan mtu to be reduced by 4 bytes. Submitted by: Doug Ambrisko (ambrisko) MFC after: 1 day
2003-04-18 17:36:13 +00:00
/*
Important fix got clobbered in the em driver, keeping VLAN HWFILTER from being used by default, this breaks stacked pseudo devices, and as it turns out, also breaks virtual machines that happen to use VLANS (didn't know that before :). Put the fix back into the em driver, and for good measure add the same code to the igb driver where it should have been anyway.
2010-04-08 00:50:43 +00:00
* Tell the upper layer(s) we
* support full VLAN capability
Tell the upper layer(s) that we support long frames. Not doing this caused the vlan mtu to be reduced by 4 bytes. Submitted by: Doug Ambrisko (ambrisko) MFC after: 1 day
2003-04-18 17:36:13 +00:00
*/
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if_setcapabilitiesbit(ifp, cap, 0);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Important fix got clobbered in the em driver, keeping VLAN HWFILTER from being used by default, this breaks stacked pseudo devices, and as it turns out, also breaks virtual machines that happen to use VLANS (didn't know that before :). Put the fix back into the em driver, and for good measure add the same code to the igb driver where it should have been anyway.
2010-04-08 00:50:43 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* Don't turn this on by default, if vlans are
* created on another pseudo device (eg. lagg)
* then vlan events are not passed thru, breaking
* operation, but with HW FILTER off it works. If
* using vlans directly on the em driver you can
* enable this and get full hardware tag filtering.
*/
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWFILTER,0);
Important fix got clobbered in the em driver, keeping VLAN HWFILTER from being used by default, this breaks stacked pseudo devices, and as it turns out, also breaks virtual machines that happen to use VLANS (didn't know that before :). Put the fix back into the em driver, and for good measure add the same code to the igb driver where it should have been anyway.
2010-04-08 00:50:43 +00:00
Change default WOL back to MAGIC only, having multicast enabled causes problems in man environments.
2010-04-28 17:37:30 +00:00
/* Enable only WOL MAGIC by default */
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
if (adapter->wol) {
igb(4) enable WOL features for this class of devices. PR: 208343 Submitted by: Kaho Tashikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-01-22 18:04:57 +00:00
if_setcapenablebit(ifp, IFCAP_WOL_MAGIC,
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
IFCAP_WOL_MCAST| IFCAP_WOL_UCAST);
igb(4) enable WOL features for this class of devices. PR: 208343 Submitted by: Kaho Tashikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-01-22 18:04:57 +00:00
} else {
if_setcapenablebit(ifp, 0, IFCAP_WOL_MAGIC |
IFCAP_WOL_MCAST| IFCAP_WOL_UCAST);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
}
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
/*
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
* Specify the media types supported by this adapter and register
* callbacks to update media and link information
*/
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
(adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
u_char fiber_type = IFM_1000_SX; /* default type */
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type == e1000_82545)
fiber_type = IFM_1000_LX;
ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
} else {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if (adapter->hw.phy.type != e1000_phy_ife) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
10/100 PHY shouldn't support gigabit media types. Submitted by: brad (brad@comstyle.com) Obtained from: OpenBSD MFC after: 1 week
2006-08-09 20:10:35 +00:00
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
Do not call voluntary panic(9) in case of if_alloc() failure. Reviewed by: jfv
2010-08-28 00:09:19 +00:00
return (0);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
- Bus DMA'fy the driver - Use htole* macros where appropriate so that the driver could work on non-x86 architectures - Use m_getcl() instead of MGETHDR/MCLGET macros Submitted by: sam (Sam Leffler)
2003-05-02 21:17:08 +00:00
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
if_softc_ctx_t scctx = adapter->shared;
int error = E1000_SUCCESS;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct em_tx_queue *que;
int i, j;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
MPASS(adapter->tx_num_queues > 0);
MPASS(adapter->tx_num_queues == ntxqsets);
/* First allocate the top level queue structs */
if (!(adapter->tx_queues =
(struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
return(ENOMEM);
}
for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* Set up some basics */
struct tx_ring *txr = &que->txr;
txr->adapter = que->adapter = adapter;
que->me = txr->me = i;
/* Allocate report status array */
if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
error = ENOMEM;
goto fail;
}
for (j = 0; j < scctx->isc_ntxd[0]; j++)
txr->tx_rsq[j] = QIDX_INVALID;
/* get the virtual and physical address of the hardware queues */
txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
txr->tx_paddr = paddrs[i*ntxqs];
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
}
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
device_printf(iflib_get_dev(ctx), "allocated for %d tx_queues\n", adapter->tx_num_queues);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (0);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
fail:
em_if_queues_free(ctx);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (error);
- Bus DMA'fy the driver - Use htole* macros where appropriate so that the driver could work on non-x86 architectures - Use m_getcl() instead of MGETHDR/MCLGET macros Submitted by: sam (Sam Leffler)
2003-05-02 21:17:08 +00:00
}
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
static int
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
int error = E1000_SUCCESS;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct em_rx_queue *que;
int i;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
MPASS(adapter->rx_num_queues > 0);
MPASS(adapter->rx_num_queues == nrxqsets);
/* First allocate the top level queue structs */
if (!(adapter->rx_queues =
(struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
error = ENOMEM;
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
goto fail;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Set up some basics */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct rx_ring *rxr = &que->rxr;
rxr->adapter = que->adapter = adapter;
rxr->que = que;
que->me = rxr->me = i;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* get the virtual and physical address of the hardware queues */
rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
rxr->rx_paddr = paddrs[i*nrxqs];
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
device_printf(iflib_get_dev(ctx), "allocated for %d rx_queues\n", adapter->rx_num_queues);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
return (0);
fail:
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
em_if_queues_free(ctx);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
return (error);
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
static void
em_if_queues_free(if_ctx_t ctx)
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
struct em_tx_queue *tx_que = adapter->tx_queues;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct em_rx_queue *rx_que = adapter->rx_queues;
busdma cleanup for em(4). - don't force busdma to pre-allocate bounce pages for parent tag. - use system supplied roundup2 macro instead of rolling its own version. - TX/RX decriptor length should be multiple of 128. There is no no need to expand the size with the multiple of 4096. - don't create/destroy DMA maps in TX/RX handlers. Use pre-allocated DMA maps. Since creating DMA maps on sparc64 is time consuming operations(resource mananger overhead), this change should boost performance on sparc64. I could get > 2x speedup on Ultra60. - TX/RX descriptors could be aligned on 128 boundary. Aligning them on PAGE_SIZE is waste of resource. - don't blindly create TX DMA tag with size of MCLBYTES * 8. The size is only valid under jumbo frame environments. Instead of using the hardcoded value, re-compute necessary size on the fly. - RX side bus_dmamap_load_mbuf_sg(9) support. - remove unused macro EM_ROUNDUP and constant EM_MMBA. Reviewed by: scottl Tested by: glebius
2005-11-21 04:17:43 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (tx_que != NULL) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
struct tx_ring *txr = &tx_que->txr;
if (txr->tx_rsq == NULL)
break;
free(txr->tx_rsq, M_DEVBUF);
txr->tx_rsq = NULL;
}
free(adapter->tx_queues, M_DEVBUF);
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
adapter->tx_queues = NULL;
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (rx_que != NULL) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
free(adapter->rx_queues, M_DEVBUF);
adapter->rx_queues = NULL;
busdma cleanup for em(4). - don't force busdma to pre-allocate bounce pages for parent tag. - use system supplied roundup2 macro instead of rolling its own version. - TX/RX decriptor length should be multiple of 128. There is no no need to expand the size with the multiple of 4096. - don't create/destroy DMA maps in TX/RX handlers. Use pre-allocated DMA maps. Since creating DMA maps on sparc64 is time consuming operations(resource mananger overhead), this change should boost performance on sparc64. I could get > 2x speedup on Ultra60. - TX/RX descriptors could be aligned on 128 boundary. Aligning them on PAGE_SIZE is waste of resource. - don't blindly create TX DMA tag with size of MCLBYTES * 8. The size is only valid under jumbo frame environments. Instead of using the hardcoded value, re-compute necessary size on the fly. - RX side bus_dmamap_load_mbuf_sg(9) support. - remove unused macro EM_ROUNDUP and constant EM_MMBA. Reviewed by: scottl Tested by: glebius
2005-11-21 04:17:43 +00:00
}
- Bus DMA'fy the driver - Use htole* macros where appropriate so that the driver could work on non-x86 architectures - Use m_getcl() instead of MGETHDR/MCLGET macros Submitted by: sam (Sam Leffler)
2003-05-02 21:17:08 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_release_hw_control(adapter);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->mta != NULL) {
free(adapter->mta, M_DEVBUF);
}
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
/*********************************************************************
*
* Enable transmit unit.
*
**********************************************************************/
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_initialize_transmit_unit(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if_softc_ctx_t scctx = adapter->shared;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct em_tx_queue *que;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct tx_ring *txr;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
struct e1000_hw *hw = &adapter->hw;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
u32 tctl, txdctl = 0, tarc, tipg = 0;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
u64 bus_addr;
caddr_t offp, endp;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
que = &adapter->tx_queues[i];
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
txr = &que->txr;
bus_addr = txr->tx_paddr;
/* Clear checksum offload context. */
offp = (caddr_t)&txr->csum_flags;
endp = (caddr_t)(txr + 1);
bzero(offp, endp - offp);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Base and Len of TX Ring */
E1000_WRITE_REG(hw, E1000_TDLEN(i),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
E1000_WRITE_REG(hw, E1000_TDBAH(i),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
(u32)(bus_addr >> 32));
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
E1000_WRITE_REG(hw, E1000_TDBAL(i),
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
(u32)bus_addr);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Init the HEAD/TAIL indices */
E1000_WRITE_REG(hw, E1000_TDT(i), 0);
E1000_WRITE_REG(hw, E1000_TDH(i), 0);
HW_DEBUGOUT2("Base = %x, Length = %x\n",
E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
txdctl = 0; /* clear txdctl */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
txdctl |= 0x1f; /* PTHRESH */
txdctl |= 1 << 8; /* HTHRESH */
txdctl |= 1 << 16;/* WTHRESH */
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
txdctl |= E1000_TXDCTL_GRAN;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
txdctl |= 1 << 25; /* LWTHRESH */
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Set the default values for the Tx Inter Packet Gap timer */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
switch (adapter->hw.mac.type) {
case e1000_80003es2lan:
tipg = DEFAULT_82543_TIPG_IPGR1;
tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
Merge in new driver from Intel, version 5.1.5. Adds support for some new chips and improves support for already supported ones. Some details, important for future merges: - if_em.c merged manually, viewing diff between new vendor driver and previous one. - if_em_hw.h dropped in from vendor, and then restored revisions 1.16, 1.17, 1.18. - if_em_hw.c dropped in from vendor, and then two liner change made, that restores support for two rare chips.
2006-04-06 17:09:03 +00:00
E1000_TIPG_IPGR2_SHIFT;
break;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
case e1000_82542:
tipg = DEFAULT_82542_TIPG_IPGT;
tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
break;
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
default:
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
(adapter->hw.phy.media_type ==
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
e1000_media_type_internal_serdes))
tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
else
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if(adapter->hw.mac.type >= e1000_82540)
E1000_WRITE_REG(&adapter->hw, E1000_TADV,
adapter->tx_abs_int_delay.value);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if ((adapter->hw.mac.type == e1000_82571) ||
(adapter->hw.mac.type == e1000_82572)) {
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
tarc |= TARC_SPEED_MODE_BIT;
E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
} else if (adapter->hw.mac.type == e1000_80003es2lan) {
/* errata: program both queues to unweighted RR */
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
tarc |= 1;
E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
tarc |= 1;
E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
} else if (adapter->hw.mac.type == e1000_82574) {
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
tarc |= TARC_ERRATA_BIT;
if ( adapter->tx_num_queues > 1) {
tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
} else
E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->tx_int_delay.value > 0)
adapter->txd_cmd |= E1000_TXD_CMD_IDE;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Program the Transmit Control Register */
tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
tctl &= ~E1000_TCTL_CT;
tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type >= e1000_82571)
tctl |= E1000_TCTL_MULR;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* This write will effectively turn on the transmit unit. */
E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (hw->mac.type == e1000_pch_spt) {
u32 reg;
reg = E1000_READ_REG(hw, E1000_IOSFPC);
reg |= E1000_RCTL_RDMTS_HEX;
E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
reg = E1000_READ_REG(hw, E1000_TARC(0));
reg |= E1000_TARC0_CB_MULTIQ_3_REQ;
E1000_WRITE_REG(hw, E1000_TARC(0), reg);
}
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
/*********************************************************************
*
* Enable receive unit.
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
*
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
**********************************************************************/
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_initialize_receive_unit(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if_softc_ctx_t scctx = adapter->shared;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct ifnet *ifp = iflib_get_ifp(ctx);
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
struct e1000_hw *hw = &adapter->hw;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct em_rx_queue *que;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
int i;
u32 rctl, rxcsum, rfctl;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
INIT_DEBUGOUT("em_initialize_receive_units: begin");
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
/*
* Make sure receives are disabled while setting
* up the descriptor ring
*/
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
rctl = E1000_READ_REG(hw, E1000_RCTL);
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
/* Do not disable if ever enabled on this hardware */
if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
Switch em(4) to the extended RX descriptor format. This matches the e1000/e1000e split in linux. Split rxbuffer and txbuffer apart to support the new RX descriptor format structures. Move rxbuffer manipulation to em_setup_rxdesc() to unify the new behavior changes. Add a RSSKEYLEN macro for help in generating the RSSKEY data structures in the card. Change em_receive_checksum() to process the new rxdescriptor format status bit. MFC after: 2 weeks Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D3447
2016-01-07 16:42:48 +00:00
/* Setup the Receive Control Register */
rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
(hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
/* Do not store bad packets */
rctl &= ~E1000_RCTL_SBP;
/* Enable Long Packet receive */
if (if_getmtu(ifp) > ETHERMTU)
rctl |= E1000_RCTL_LPE;
else
rctl &= ~E1000_RCTL_LPE;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* Strip the CRC */
if (!em_disable_crc_stripping)
Switch em(4) to the extended RX descriptor format. This matches the e1000/e1000e split in linux. Split rxbuffer and txbuffer apart to support the new RX descriptor format structures. Move rxbuffer manipulation to em_setup_rxdesc() to unify the new behavior changes. Add a RSSKEYLEN macro for help in generating the RSSKEY data structures in the card. Change em_receive_checksum() to process the new rxdescriptor format status bit. MFC after: 2 weeks Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D3447
2016-01-07 16:42:48 +00:00
rctl |= E1000_RCTL_SECRC;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->hw.mac.type >= e1000_82540) {
E1000_WRITE_REG(&adapter->hw, E1000_RADV,
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->rx_abs_int_delay.value);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*
* Set the interrupt throttling rate. Value is calculated
* as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
*/
E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
}
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
adapter->rx_int_delay.value);
- Move to array based indexing for TX/RX descriptor/buffer management - Added support for ITR (interrupt throttle register). This feature is available on adapters based on 82545 and above - Fixed problem with vlan support when traffic has priority bits set. (kern/45907) PR: kern/45907 MFC after: 1 week
2002-12-23 19:11:23 +00:00
Switch em(4) to the extended RX descriptor format. This matches the e1000/e1000e split in linux. Split rxbuffer and txbuffer apart to support the new RX descriptor format structures. Move rxbuffer manipulation to em_setup_rxdesc() to unify the new behavior changes. Add a RSSKEYLEN macro for help in generating the RSSKEY data structures in the card. Change em_receive_checksum() to process the new rxdescriptor format status bit. MFC after: 2 weeks Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D3447
2016-01-07 16:42:48 +00:00
/* Use extended rx descriptor formats */
rfctl = E1000_READ_REG(hw, E1000_RFCTL);
rfctl |= E1000_RFCTL_EXTEN;
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* When using MSIX interrupts we need to throttle
* using the EITR register (82574 only)
*/
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
if (hw->mac.type == e1000_82574) {
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
for (int i = 0; i < 4; i++)
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
E1000_WRITE_REG(hw, E1000_EITR_82574(i),
DEFAULT_ITR);
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
/* Disable accelerated acknowledge */
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
rfctl |= E1000_RFCTL_ACK_DIS;
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
}
Switch em(4) to the extended RX descriptor format. This matches the e1000/e1000e split in linux. Split rxbuffer and txbuffer apart to support the new RX descriptor format structures. Move rxbuffer manipulation to em_setup_rxdesc() to unify the new behavior changes. Add a RSSKEYLEN macro for help in generating the RSSKEY data structures in the card. Change em_receive_checksum() to process the new rxdescriptor format status bit. MFC after: 2 weeks Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D3447
2016-01-07 16:42:48 +00:00
E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
Corrections to the RX checksum code, make sure its disabled as well as enabled when necessary. And simplify the checksum routine itself, adding UDP bit to the test. Thanks to Kevin Lo for pointing out the problems and code suggestions.
2013-04-15 17:01:42 +00:00
rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
adapter->hw.mac.type >= e1000_82543) {
if (adapter->tx_num_queues > 1) {
if (adapter->hw.mac.type >= igb_mac_min) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
rxcsum |= E1000_RXCSUM_PCSD;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (hw->mac.type != e1000_82575)
rxcsum |= E1000_RXCSUM_CRCOFL;
} else
rxcsum |= E1000_RXCSUM_TUOFL |
E1000_RXCSUM_IPOFL |
E1000_RXCSUM_PCSD;
} else {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (adapter->hw.mac.type >= igb_mac_min)
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
rxcsum |= E1000_RXCSUM_IPPCSE;
else
rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
if (adapter->hw.mac.type > e1000_82575)
rxcsum |= E1000_RXCSUM_CRCOFL;
}
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
} else
Corrections to the RX checksum code, make sure its disabled as well as enabled when necessary. And simplify the checksum routine itself, adding UDP bit to the test. Thanks to Kevin Lo for pointing out the problems and code suggestions.
2013-04-15 17:01:42 +00:00
rxcsum &= ~E1000_RXCSUM_TUOFL;
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Corrections to the RX checksum code, make sure its disabled as well as enabled when necessary. And simplify the checksum routine itself, adding UDP bit to the test. Thanks to Kevin Lo for pointing out the problems and code suggestions.
2013-04-15 17:01:42 +00:00
E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->rx_num_queues > 1) {
if (adapter->hw.mac.type >= igb_mac_min)
igb_initialize_rss_mapping(adapter);
else
em_initialize_rss_mapping(adapter);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* XXX TEMPORARY WORKAROUND: on some systems with 82573
* long latencies are observed, like Lenovo X60. This
* change eliminates the problem, but since having positive
* values in RDTR is a known source of problems on other
* platforms another solution is being sought.
*/
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
if (hw->mac.type == e1000_82573)
E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct rx_ring *rxr = &que->rxr;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Setup the Base and Length of the Rx Descriptor Ring */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
u64 bus_addr = rxr->rx_paddr;
#if 0
u32 rdt = adapter->rx_num_queues -1; /* default */
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
#endif
simplify the code to initialize the RDT while in netmap mode.
2013-05-09 16:57:02 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
E1000_WRITE_REG(hw, E1000_RDLEN(i),
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
/* Setup the Head and Tail Descriptor Pointers */
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
E1000_WRITE_REG(hw, E1000_RDH(i), 0);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
E1000_WRITE_REG(hw, E1000_RDT(i), 0);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
/*
* Set PTHRESH for improved jumbo performance
* According to 10.2.5.11 of Intel 82574 Datasheet,
* RXDCTL(1) is written whenever RXDCTL(0) is written.
* Only write to RXDCTL(1) if there is a need for different
* settings.
*/
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
if (((adapter->hw.mac.type == e1000_ich9lan) ||
(adapter->hw.mac.type == e1000_pch2lan) ||
(adapter->hw.mac.type == e1000_ich10lan)) &&
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
(if_getmtu(ifp) > ETHERMTU)) {
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
Switch em(4) to the extended RX descriptor format. This matches the e1000/e1000e split in linux. Split rxbuffer and txbuffer apart to support the new RX descriptor format structures. Move rxbuffer manipulation to em_setup_rxdesc() to unify the new behavior changes. Add a RSSKEYLEN macro for help in generating the RSSKEY data structures in the card. Change em_receive_checksum() to process the new rxdescriptor format status bit. MFC after: 2 weeks Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D3447
2016-01-07 16:42:48 +00:00
} else if (adapter->hw.mac.type == e1000_82574) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (int i = 0; i < adapter->rx_num_queues; i++) {
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
Switch em(4) to the extended RX descriptor format. This matches the e1000/e1000e split in linux. Split rxbuffer and txbuffer apart to support the new RX descriptor format structures. Move rxbuffer manipulation to em_setup_rxdesc() to unify the new behavior changes. Add a RSSKEYLEN macro for help in generating the RSSKEY data structures in the card. Change em_receive_checksum() to process the new rxdescriptor format status bit. MFC after: 2 weeks Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D3447
2016-01-07 16:42:48 +00:00
rxdctl |= 0x20; /* PTHRESH */
rxdctl |= 4 << 8; /* HTHRESH */
rxdctl |= 4 << 16;/* WTHRESH */
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
rxdctl |= 1 << 24; /* Switch to granularity */
E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
}
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
} else if (adapter->hw.mac.type >= igb_mac_min) {
u32 psize, srrctl = 0;
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
if (if_getmtu(ifp) > ETHERMTU) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* Set maximum packet len */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (adapter->rx_mbuf_sz <= 4096) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
} else if (adapter->rx_mbuf_sz > 4096) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
psize = scctx->isc_max_frame_size;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/* are we on a vlan? */
if (ifp->if_vlantrunk != NULL)
psize += VLAN_TAG_SIZE;
E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
} else {
srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
rctl |= E1000_RCTL_SZ_2048;
}
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
/*
* If TX flow control is disabled and there's >1 queue defined,
* enable DROP.
*
* This drops frames rather than hanging the RX MAC for all queues.
*/
if ((adapter->rx_num_queues > 1) &&
(adapter->fc == e1000_fc_none ||
adapter->fc == e1000_fc_rx_pause)) {
srrctl |= E1000_SRRCTL_DROP_EN;
}
/* Setup the Base and Length of the Rx Descriptor Rings */
for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
struct rx_ring *rxr = &que->rxr;
u64 bus_addr = rxr->rx_paddr;
u32 rxdctl;
#ifdef notyet
/* Configure for header split? -- ignore for now */
rxr->hdr_split = igb_header_split;
#else
srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
#endif
E1000_WRITE_REG(hw, E1000_RDLEN(i),
scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
E1000_WRITE_REG(hw, E1000_RDBAH(i),
(uint32_t)(bus_addr >> 32));
E1000_WRITE_REG(hw, E1000_RDBAL(i),
(uint32_t)bus_addr);
E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
/* Enable this Queue */
rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
rxdctl &= 0xFFF00000;
rxdctl |= IGB_RX_PTHRESH;
rxdctl |= IGB_RX_HTHRESH << 8;
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
rxdctl |= IGB_RX_WTHRESH << 16;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
}
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
} else if (adapter->hw.mac.type >= e1000_pch2lan) {
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if (if_getmtu(ifp) > ETHERMTU)
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
else
e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* Make sure VLAN Filters are off */
rctl &= ~E1000_RCTL_VFE;
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
Restore PBA setup for igb(4) class devices. Do no write to PBA register on igb(4) devices unless we need to make adjustments for the 82575 and jumbo frames. Remove redundant LPE/~LPE assignments. Move e1000_lv_jumbo_workaround_ich8lan() invokcation into a block so that its not executed in the igb case. Move em(4) class assignments of RCTL values to its own code block. Adjust a few direct accesses of ifp->mtu to use accessor functions. PR: 216734 Submitted by: Kaho Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-02-19 05:06:29 +00:00
if (adapter->hw.mac.type < igb_mac_min) {
if (adapter->rx_mbuf_sz == MCLBYTES)
rctl |= E1000_RCTL_SZ_2048;
else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
/* ensure we clear use DTYPE of 00 here */
rctl &= ~0x00000C00;
}
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
/* Write out the settings */
E1000_WRITE_REG(hw, E1000_RCTL, rctl);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
return;
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_vlan_register(if_ctx_t ctx, u16 vtag)
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
u32 index, bit;
This delta is primarily a fix for es2lan devices that will sometimes fail to initialize problem due to a lock contention with management hardware. However, in order to deliver that fix it was necessary to take a shared code update as a whole, and this required scattered changes in the core code to be compatible. The em driver now has VLAN HW support added as the igb driver had previously. MFC after: ASAP - in time for 7.1 RELEASE
2008-11-26 23:57:23 +00:00
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
index = (vtag >> 5) & 0x7F;
bit = vtag & 0x1F;
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
adapter->shadow_vfta[index] |= (1 << bit);
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
++adapter->num_vlans;
This delta is primarily a fix for es2lan devices that will sometimes fail to initialize problem due to a lock contention with management hardware. However, in order to deliver that fix it was necessary to take a shared code update as a whole, and this required scattered changes in the core code to be compatible. The em driver now has VLAN HW support added as the igb driver had previously. MFC after: ASAP - in time for 7.1 RELEASE
2008-11-26 23:57:23 +00:00
}
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
This delta is primarily a fix for es2lan devices that will sometimes fail to initialize problem due to a lock contention with management hardware. However, in order to deliver that fix it was necessary to take a shared code update as a whole, and this required scattered changes in the core code to be compatible. The em driver now has VLAN HW support added as the igb driver had previously. MFC after: ASAP - in time for 7.1 RELEASE
2008-11-26 23:57:23 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
u32 index, bit;
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
index = (vtag >> 5) & 0x7F;
bit = vtag & 0x1F;
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
adapter->shadow_vfta[index] &= ~(1 << bit);
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
--adapter->num_vlans;
}
static void
em_setup_vlan_hw_support(struct adapter *adapter)
{
struct e1000_hw *hw = &adapter->hw;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
u32 reg;
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* We get here thru init_locked, meaning
* a soft reset, this has already cleared
* the VFTA and other state, so if there
* have been no vlan's registered do nothing.
*/
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
if (adapter->num_vlans == 0)
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return;
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* A soft reset zero's out the VFTA, so
* we need to repopulate it now.
*/
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
for (int i = 0; i < EM_VFTA_SIZE; i++)
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
if (adapter->shadow_vfta[i] != 0)
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
i, adapter->shadow_vfta[i]);
Updates for both the em and igb drivers, add support for multiqueue tx, shared code updates, new device support, and some bug fixes.
2009-06-24 17:41:29 +00:00
reg = E1000_READ_REG(hw, E1000_CTRL);
reg |= E1000_CTRL_VME;
E1000_WRITE_REG(hw, E1000_CTRL, reg);
/* Enable the Filter Table */
reg = E1000_READ_REG(hw, E1000_RCTL);
reg &= ~E1000_RCTL_CFIEN;
reg |= E1000_RCTL_VFE;
E1000_WRITE_REG(hw, E1000_RCTL, reg);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_enable_intr(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
struct e1000_hw *hw = &adapter->hw;
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
u32 ims_mask = IMS_ENABLE_MASK;
Correct broken build.
2010-04-10 07:26:51 +00:00
if (hw->mac.type == e1000_82574) {
Reset the EIAC register to include the LINK status bit and restore link up/down notifications. Submitted by: Franco Fichtner <franco@opnsense.org>
2017-01-12 14:28:32 +00:00
E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
Increase EM_MAX_SCATTER to 64 such that the size of em_xmit()::segs[EM_MAX_SCATTER] doesn't get overrun by things like NFS that can and do shove more than 32 segs when being used with em(4) and TSO4. Update tso handling code in em_xmit() with update from jhb@ in email thread: https://lists.freebsd.org/pipermail/freebsd-net/2014-July/039306.html set ifp->if_hw_tsomax, ifp->if_hw_tsomaxsegcount & ifp->if_hw_tsomaxsegsize to appropriate values. Define a TSO workaround "magic" number of 4 that is used to avoid an alignment issue in hardware. Change a couple of integer values that were used as booleans to actual bool types. Ensure that em_enable_intr() enables the appropriate mask of interrupts and not just a hardcoded define of values. PR: 200221 199174 195078 Differential Revision: https://reviews.freebsd.org/D3192 Reviewed by: erj jhb hiren MFC after: 2 weeks Sponsored by: Limelight Networks
2015-08-16 19:43:44 +00:00
ims_mask |= adapter->ims;
Add missing 'else' to conditional. This doesn't really affect the code flow or configuration in any way.
2017-03-24 14:27:29 +00:00
} else if (adapter->intr_type == IFLIB_INTR_MSIX && hw->mac.type >= igb_mac_min) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
u32 mask = (adapter->que_mask | adapter->link_mask);
E1000_WRITE_REG(&adapter->hw, E1000_EIAC, mask);
E1000_WRITE_REG(&adapter->hw, E1000_EIAM, mask);
E1000_WRITE_REG(&adapter->hw, E1000_EIMS, mask);
ims_mask = E1000_IMS_LSC;
}
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_disable_intr(if_ctx_t ctx)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
struct e1000_hw *hw = &adapter->hw;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (adapter->intr_type == IFLIB_INTR_MSIX) {
if (hw->mac.type >= igb_mac_min)
E1000_WRITE_REG(&adapter->hw, E1000_EIMC, ~0);
E1000_WRITE_REG(&adapter->hw, E1000_EIAC, 0);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
}
Corrected a workaround that should only be applied to one adapter. Workaround was causing device hangs when incorrectly applied to other adapters. PR: kern/66634
2005-01-01 19:57:23 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/*
* Bit of a misnomer, what this really means is
* to enable OS management of the system... aka
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
* to disable special hardware management features
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
*/
static void
em_init_manageability(struct adapter *adapter)
{
/* A shared code workaround */
#define E1000_82542_MANC2H E1000_MANC2H
if (adapter->has_manage) {
int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
/* disable hardware interception of ARP */
manc &= ~(E1000_MANC_ARP_EN);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* enable receiving management packets to the host */
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
manc |= E1000_MANC_EN_MNG2HOST;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
#define E1000_MNG2HOST_PORT_623 (1 << 5)
#define E1000_MNG2HOST_PORT_664 (1 << 6)
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
manc2h |= E1000_MNG2HOST_PORT_623;
manc2h |= E1000_MNG2HOST_PORT_664;
E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
}
}
/*
* Give control back to hardware management
* controller if there is one.
*/
static void
em_release_manageability(struct adapter *adapter)
{
if (adapter->has_manage) {
int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
/* re-enable hardware interception of ARP */
manc |= E1000_MANC_ARP_EN;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
manc &= ~E1000_MANC_EN_MNG2HOST;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
}
}
/*
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
* em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
* For ASF and Pass Through versions of f/w this means
* that the driver is loaded. For AMT version type f/w
* this means that the network i/f is open.
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
*/
static void
em_get_hw_control(struct adapter *adapter)
{
u32 ctrl_ext, swsm;
Restore igb(4) code dropped during iflib conversion - restore newer code for vf, i350, i210, i211 - restore dmac init code for i354 and i350 - restore WUC/WUFC update - check for igb mac type before attempting trying to assert a media changed event. - handle link events for igb(4) and em(4) devices differently and appropriately for their respective model types. Submitted by: Matt Macy <mmacy@mattmacy.io> Sponsored by: Limelight Networks
2017-07-19 22:41:22 +00:00
if (adapter->vf_ifp)
return;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
if (adapter->hw.mac.type == e1000_82573) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
swsm | E1000_SWSM_DRV_LOAD);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
return;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
}
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
/* else */
ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
}
/*
* em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
* For ASF and Pass Through versions of f/w this means that
* the driver is no longer loaded. For AMT versions of the
* f/w this means that the network i/f is closed.
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
*/
static void
em_release_hw_control(struct adapter *adapter)
{
u32 ctrl_ext, swsm;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
if (!adapter->has_manage)
return;
if (adapter->hw.mac.type == e1000_82573) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
swsm & ~E1000_SWSM_DRV_LOAD);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
return;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
}
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
/* else */
ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
return;
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
static int
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
em_is_valid_ether_addr(u8 *addr)
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
{
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
return (FALSE);
}
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (TRUE);
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
}
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
/*
** Parse the interface capabilities with regard
** to both system management and wake-on-lan for
** later use.
*/
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_get_wakeup(if_ctx_t ctx)
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
device_t dev = iflib_get_dev(ctx);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
u16 eeprom_data = 0, device_id, apme_mask;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
apme_mask = EM_EEPROM_APME;
switch (adapter->hw.mac.type) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
case e1000_82542:
case e1000_82543:
break;
case e1000_82544:
e1000_read_nvm(&adapter->hw,
NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
apme_mask = EM_82544_APME;
break;
case e1000_82546:
case e1000_82546_rev_3:
if (adapter->hw.bus.func == 1) {
e1000_read_nvm(&adapter->hw,
NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
break;
} else
e1000_read_nvm(&adapter->hw,
NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
break;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
case e1000_82573:
case e1000_82583:
adapter->has_amt = TRUE;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* FALLTHROUGH */
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
case e1000_82571:
case e1000_82572:
case e1000_80003es2lan:
if (adapter->hw.bus.func == 1) {
e1000_read_nvm(&adapter->hw,
NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
break;
} else
e1000_read_nvm(&adapter->hw,
NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
break;
case e1000_ich8lan:
case e1000_ich9lan:
case e1000_ich10lan:
case e1000_pchlan:
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
case e1000_pch2lan:
Restore fixup for newer em(4) devices WOL capabilities post iflib integration. PR: 208343
2017-01-12 14:47:44 +00:00
case e1000_pch_lpt:
case e1000_pch_spt:
igb(4) enable WOL features for this class of devices. PR: 208343 Submitted by: Kaho Tashikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-01-22 18:04:57 +00:00
case e1000_82575: /* listing all igb devices */
case e1000_82576:
case e1000_82580:
case e1000_i350:
case e1000_i354:
case e1000_i210:
case e1000_i211:
case e1000_vfadapt:
case e1000_vfadapt_i350:
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
apme_mask = E1000_WUC_APME;
adapter->has_amt = TRUE;
eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
break;
default:
e1000_read_nvm(&adapter->hw,
NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
break;
}
if (eeprom_data & apme_mask)
adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* We have the eeprom settings, now apply the special cases
* where the eeprom may be wrong or the board won't support
* wake on lan on a particular port
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
*/
device_id = pci_get_device(dev);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
switch (device_id) {
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
case E1000_DEV_ID_82546GB_PCIE:
adapter->wol = 0;
break;
case E1000_DEV_ID_82546EB_FIBER:
case E1000_DEV_ID_82546GB_FIBER:
/* Wake events only supported on port A for dual fiber
* regardless of eeprom setting */
if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
E1000_STATUS_FUNC_1)
adapter->wol = 0;
break;
case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* if quad port adapter, disable WoL on all but port A */
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
if (global_quad_port_a != 0)
adapter->wol = 0;
/* Reset for multiple quad port adapters */
if (++global_quad_port_a == 4)
global_quad_port_a = 0;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
break;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
case E1000_DEV_ID_82571EB_FIBER:
/* Wake events only supported on port A for dual fiber
* regardless of eeprom setting */
if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
E1000_STATUS_FUNC_1)
adapter->wol = 0;
break;
case E1000_DEV_ID_82571EB_QUAD_COPPER:
case E1000_DEV_ID_82571EB_QUAD_FIBER:
case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
/* if quad port adapter, disable WoL on all but port A */
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
if (global_quad_port_a != 0)
adapter->wol = 0;
/* Reset for multiple quad port adapters */
if (++global_quad_port_a == 4)
global_quad_port_a = 0;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
break;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
}
return;
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/*
* Enable PCI Wake On Lan capability
*/
Hook the identification LEDs of igb(4), lem(4) and em(4) devices up with led(4) so they can be lit or f.e. made blink via `echo f2 > /dev/led/em0` for localization purposes. Approved by: jfv MFC afer: 1 week (after r205869)
2010-03-31 20:43:24 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_enable_wakeup(if_ctx_t ctx)
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
device_t dev = iflib_get_dev(ctx);
if_t ifp = iflib_get_ifp(ctx);
Don't set any WOL enabling hardware bits if WOL isn't requested according to the enabled interface capability bits. Also remove some dead code, which tried to preserve already set contents of E1000_WUC while that register is completely overwritten shortly after in all cases.
2017-08-28 22:09:12 +00:00
int error = 0;
u32 pmc, ctrl, ctrl_ext, rctl;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
u16 status;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
Don't set any WOL enabling hardware bits if WOL isn't requested according to the enabled interface capability bits. Also remove some dead code, which tried to preserve already set contents of E1000_WUC while that register is completely overwritten shortly after in all cases.
2017-08-28 22:09:12 +00:00
if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
return;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* Determine type of Wakeup: note that wol
* is set with all bits on by default.
*/
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
adapter->wol &= ~E1000_WUFC_MAG;
igb(4) enable WOL features for this class of devices. PR: 208343 Submitted by: Kaho Tashikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-01-22 18:04:57 +00:00
if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
adapter->wol &= ~E1000_WUFC_EX;
Convert em(4) to use the driver API. Submitted by: Anuranjan Shukla <anshukla@juniper.net> Obtained from: Juniper Networks, Inc.
2014-06-02 18:52:03 +00:00
if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
adapter->wol &= ~E1000_WUFC_MC;
else {
rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
rctl |= E1000_RCTL_MPE;
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
}
Don't set any WOL enabling hardware bits if WOL isn't requested according to the enabled interface capability bits. Also remove some dead code, which tried to preserve already set contents of E1000_WUC while that register is completely overwritten shortly after in all cases.
2017-08-28 22:09:12 +00:00
if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
goto pme;
/* Advertise the wakeup capability */
ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
/* Keep the laser running on Fiber adapters */
if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
}
if ((adapter->hw.mac.type == e1000_ich8lan) ||
(adapter->hw.mac.type == e1000_pchlan) ||
(adapter->hw.mac.type == e1000_ich9lan) ||
(adapter->hw.mac.type == e1000_ich10lan))
e1000_suspend_workarounds_ich8lan(&adapter->hw);
igb(4) enable WOL features for this class of devices. PR: 208343 Submitted by: Kaho Tashikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-01-22 18:04:57 +00:00
if ( adapter->hw.mac.type >= e1000_pchlan) {
Don't set any WOL enabling hardware bits if WOL isn't requested according to the enabled interface capability bits. Also remove some dead code, which tried to preserve already set contents of E1000_WUC while that register is completely overwritten shortly after in all cases.
2017-08-28 22:09:12 +00:00
error = em_enable_phy_wakeup(adapter);
if (error)
goto pme;
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
} else {
Don't set any WOL enabling hardware bits if WOL isn't requested according to the enabled interface capability bits. Also remove some dead code, which tried to preserve already set contents of E1000_WUC while that register is completely overwritten shortly after in all cases.
2017-08-28 22:09:12 +00:00
/* Enable wakeup by the MAC */
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
}
if (adapter->hw.phy.type == e1000_phy_igp_3)
e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
Don't set any WOL enabling hardware bits if WOL isn't requested according to the enabled interface capability bits. Also remove some dead code, which tried to preserve already set contents of E1000_WUC while that register is completely overwritten shortly after in all cases.
2017-08-28 22:09:12 +00:00
pme:
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
Don't set any WOL enabling hardware bits if WOL isn't requested according to the enabled interface capability bits. Also remove some dead code, which tried to preserve already set contents of E1000_WUC while that register is completely overwritten shortly after in all cases.
2017-08-28 22:09:12 +00:00
if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
return;
}
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* WOL in the newer chipset interfaces (pchlan)
* require thing to be copied into the phy
*/
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
static int
em_enable_phy_wakeup(struct adapter *adapter)
{
struct e1000_hw *hw = &adapter->hw;
u32 mreg, ret = 0;
u16 preg;
/* copy MAC RARs to PHY RARs */
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
e1000_copy_rx_addrs_to_phy_ich8lan(hw);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
/* copy MAC MTA to PHY MTA */
for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
e1000_write_phy_reg(hw, BM_MTA(i) + 1,
(u16)((mreg >> 16) & 0xFFFF));
}
/* configure PHY Rx Control register */
e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
mreg = E1000_READ_REG(hw, E1000_RCTL);
if (mreg & E1000_RCTL_UPE)
preg |= BM_RCTL_UPE;
if (mreg & E1000_RCTL_MPE)
preg |= BM_RCTL_MPE;
preg &= ~(BM_RCTL_MO_MASK);
if (mreg & E1000_RCTL_MO_3)
preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
<< BM_RCTL_MO_SHIFT);
if (mreg & E1000_RCTL_BAM)
preg |= BM_RCTL_BAM;
if (mreg & E1000_RCTL_PMCF)
preg |= BM_RCTL_PMCF;
mreg = E1000_READ_REG(hw, E1000_CTRL);
if (mreg & E1000_CTRL_RFCE)
preg |= BM_RCTL_RFCE;
e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
/* enable PHY wakeup in MAC register */
E1000_WRITE_REG(hw, E1000_WUC,
igb(4) enable WOL features for this class of devices. PR: 208343 Submitted by: Kaho Tashikazu <kaho@elam.kais.kyoto-u.ac.jp>
2017-01-22 18:04:57 +00:00
E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
Resync with Intel versions of both the em and igb drivers. These add new hardware support, most importantly the pch (i5 chipset) in the em driver. Also, both drivers now have the simplified (and I hope improved) watchdog code. The igb driver uses the new RX cleanup that I first implemented in ixgbe. em - version 6.9.24 igb - version 1.8.4
2009-12-08 01:07:44 +00:00
E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
/* configure and enable PHY wakeup in PHY registers */
e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
/* activate PHY wakeup */
ret = hw->phy.ops.acquire(hw);
if (ret) {
printf("Could not acquire PHY\n");
return ret;
}
e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
(BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
if (ret) {
printf("Could not read PHY page 769\n");
goto out;
}
preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
if (ret)
printf("Could not set PHY Host Wakeup bit\n");
out:
hw->phy.ops.release(hw);
return ret;
}
Hook the identification LEDs of igb(4), lem(4) and em(4) devices up with led(4) so they can be lit or f.e. made blink via `echo f2 > /dev/led/em0` for localization purposes. Approved by: jfv MFC afer: 1 week (after r205869)
2010-03-31 20:43:24 +00:00
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_led_func(if_ctx_t ctx, int onoff)
Hook the identification LEDs of igb(4), lem(4) and em(4) devices up with led(4) so they can be lit or f.e. made blink via `echo f2 > /dev/led/em0` for localization purposes. Approved by: jfv MFC afer: 1 week (after r205869)
2010-03-31 20:43:24 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
Hook the identification LEDs of igb(4), lem(4) and em(4) devices up with led(4) so they can be lit or f.e. made blink via `echo f2 > /dev/led/em0` for localization purposes. Approved by: jfv MFC afer: 1 week (after r205869)
2010-03-31 20:43:24 +00:00
if (onoff) {
e1000_setup_led(&adapter->hw);
e1000_led_on(&adapter->hw);
} else {
e1000_led_off(&adapter->hw);
e1000_cleanup_led(&adapter->hw);
}
}
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* Disable the L0S and L1 LINK states
*/
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
static void
em_disable_aspm(struct adapter *adapter)
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
int base, reg;
u16 link_cap,link_ctrl;
device_t dev = adapter->dev;
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
switch (adapter->hw.mac.type) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
case e1000_82573:
case e1000_82574:
case e1000_82583:
break;
default:
return;
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
}
Do a sweep of the tree replacing calls to pci_find_extcap() with calls to pci_find_cap() instead.
2011-03-23 13:10:15 +00:00
if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
return;
Align the PCI Express #defines with the style used for the PCI-X #defines. This also has the advantage that it makes the names more compact, iand also allows us to correct the non-uniform naming of the PCIM_LINK_* defines, making them all consistent amongst themselves. This is a mostly mechanical rename: s/PCIR_EXPRESS_/PCIER_/g s/PCIM_EXP_/PCIEM_/g s/PCIM_LINK_/PCIEM_LINK_/g When this is MFC'd, #defines will be added for the old names to assist out-of-tree drivers. Discussed with: jhb MFC after: 1 week
2012-09-18 22:04:59 +00:00
reg = base + PCIER_LINK_CAP;
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
link_cap = pci_read_config(dev, reg, 2);
Align the PCI Express #defines with the style used for the PCI-X #defines. This also has the advantage that it makes the names more compact, iand also allows us to correct the non-uniform naming of the PCIM_LINK_* defines, making them all consistent amongst themselves. This is a mostly mechanical rename: s/PCIR_EXPRESS_/PCIER_/g s/PCIM_EXP_/PCIEM_/g s/PCIM_LINK_/PCIEM_LINK_/g When this is MFC'd, #defines will be added for the old names to assist out-of-tree drivers. Discussed with: jhb MFC after: 1 week
2012-09-18 22:04:59 +00:00
if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
return;
Align the PCI Express #defines with the style used for the PCI-X #defines. This also has the advantage that it makes the names more compact, iand also allows us to correct the non-uniform naming of the PCIM_LINK_* defines, making them all consistent amongst themselves. This is a mostly mechanical rename: s/PCIR_EXPRESS_/PCIER_/g s/PCIM_EXP_/PCIEM_/g s/PCIM_LINK_/PCIEM_LINK_/g When this is MFC'd, #defines will be added for the old names to assist out-of-tree drivers. Discussed with: jhb MFC after: 1 week
2012-09-18 22:04:59 +00:00
reg = base + PCIER_LINK_CTL;
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
link_ctrl = pci_read_config(dev, reg, 2);
Switch some PCI register reads from using magic numbers to using the names defined in pcireg.h MFC after: 1 week
2012-09-19 12:27:23 +00:00
link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
The purpose of this change is to add a routine to disable ASPM L0S and L1 LINK states on 82573, 82574, and 82583. The theory is that this is behind certain hangs being experienced by some customers. Also included a small optimization in the rxeof routine that was in my internal code. Change the PBA size for pchlan, it was incorrect. MFC after: 3 days
2010-11-24 22:24:07 +00:00
pci_write_config(dev, reg, link_ctrl, 2);
return;
}
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
/**********************************************************************
*
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
* Update the board statistics counters.
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*
**********************************************************************/
static void
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_update_stats_counters(struct adapter *adapter)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
if(adapter->hw.phy.media_type == e1000_media_type_copper ||
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
(E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
Simplify hang detection by stealing the techniques used in ixl(4) and applying them to em(4). Rely on iterations through the local timer, and the tx queue state to determine if an actual hang has occurred. Any time a descriptor is used (packet sent), the tx queue is flagged as busy. Then when txeof runs, it either clears the flag when all is clean, or resets it to 1 if ANY are cleaned, if nothing is cleaned it increments the flag. Local timer simply checks to see if busy ever reaches MAX (10, which is compile time configurable), and then sets it as HUNG, at that point there is one more timer cycle in which to have any cleans, if not a watchdog reset will occur. Differential Revision: https://reviews.freebsd.org/D2019 Submitted by: jfv Reviewed by: hiren Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks
2015-06-02 18:28:41 +00:00
adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
Move pause frame counter out of struct if_ctx and into struct if_softc_ctx_t so that we can use it in iflib to detect pause frames. The igb(4) driver definitely used to use this in its old timer function and I see no reason to restrict it to that driver only. Sponsored by: Limelight Networks
2017-04-07 00:33:03 +00:00
/*
** For watchdog management we need to know if we have been
** paused during the last interval, so capture that here.
*/
adapter->shared->isc_pause_frames = adapter->stats.xoffrxc;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* For the 64-bit byte counters the low dword must be read first. */
/* Both registers clear on the read of the high dword */
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
/* Interrupt Counts */
adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if (adapter->hw.mac.type >= e1000_82543) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->stats.algnerrc +=
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->stats.rxerrc +=
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_READ_REG(&adapter->hw, E1000_RXERRC);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->stats.tncrs +=
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_READ_REG(&adapter->hw, E1000_TNCRS);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->stats.cexterr +=
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->stats.tsctc +=
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_READ_REG(&adapter->hw, E1000_TSCTC);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->stats.tsctfc +=
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
}
- Use if_inc_counter() to increment various counters. - Do not ever set a counter to a value. For those counters that we don't increment, but return directly from hardware create cases in if_get_counter() method. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-18 15:56:14 +00:00
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
- Use if_inc_counter() to increment various counters. - Do not ever set a counter to a value. For those counters that we don't increment, but return directly from hardware create cases in if_get_counter() method. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-18 15:56:14 +00:00
static uint64_t
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
- Use if_inc_counter() to increment various counters. - Do not ever set a counter to a value. For those counters that we don't increment, but return directly from hardware create cases in if_get_counter() method. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-18 15:56:14 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
struct ifnet *ifp = iflib_get_ifp(ctx);
- Use if_inc_counter() to increment various counters. - Do not ever set a counter to a value. For those counters that we don't increment, but return directly from hardware create cases in if_get_counter() method. Sponsored by: Netflix Sponsored by: Nginx, Inc.
2014-09-18 15:56:14 +00:00
switch (cnt) {
case IFCOUNTER_COLLISIONS:
return (adapter->stats.colc);
case IFCOUNTER_IERRORS:
return (adapter->dropped_pkts + adapter->stats.rxerrc +
adapter->stats.crcerrs + adapter->stats.algnerrc +
adapter->stats.ruc + adapter->stats.roc +
adapter->stats.mpc + adapter->stats.cexterr);
case IFCOUNTER_OERRORS:
return (adapter->stats.ecol + adapter->stats.latecol +
adapter->watchdog_events);
default:
return (if_get_counter_default(ifp, cnt));
}
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
/* Export a single 32-bit register via a read-only sysctl. */
static int
em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter;
u_int val;
adapter = oidp->oid_arg1;
val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
return (sysctl_handle_int(oidp, &val, 0, req));
}
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
/*
* Add sysctl variables, one per statistic, to the system.
*/
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
static void
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
em_add_hw_stats(struct adapter *adapter)
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
device_t dev = iflib_get_dev(adapter->ctx);
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
struct em_tx_queue *tx_que = adapter->tx_queues;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct em_rx_queue *rx_que = adapter->rx_queues;
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
struct sysctl_oid *tree = device_get_sysctl_tree(dev);
struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
struct e1000_hw_stats *stats = &adapter->stats;
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
struct sysctl_oid *stat_node, *queue_node, *int_node;
struct sysctl_oid_list *stat_list, *queue_list, *int_list;
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
#define QUEUE_NAME_LEN 32
char namebuf[QUEUE_NAME_LEN];
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
/* Driver Statistics */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->dropped_pkts,
"Driver dropped packets");
Given that em(4), lem(4) and igb(4) hardware doesn't require the alignment guarantees provided by m_defrag(9), use m_collapse(9) instead for performance reasons. While at it, sanitize the statistics softc members, i. e. retire unused ones and add SYSCTL nodes missing for actually used ones. Differential Revision: https://reviews.freebsd.org/D4717
2016-01-13 21:47:27 +00:00
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
CTLFLAG_RD, &adapter->link_irq,
"Link MSIX IRQ Handled");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_fail",
Given that em(4), lem(4) and igb(4) hardware doesn't require the alignment guarantees provided by m_defrag(9), use m_collapse(9) instead for performance reasons. While at it, sanitize the statistics softc members, i. e. retire unused ones and add SYSCTL nodes missing for actually used ones. Differential Revision: https://reviews.freebsd.org/D4717
2016-01-13 21:47:27 +00:00
CTLFLAG_RD, &adapter->mbuf_defrag_failed,
"Defragmenting mbuf chain failed");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->no_tx_dma_setup,
"Driver tx dma failure in xmit");
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
CTLFLAG_RD, &adapter->rx_overruns,
"RX overruns");
SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
CTLFLAG_RD, &adapter->watchdog_events,
"Watchdog timeouts");
SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
em_sysctl_reg_handler, "IU",
"Device Control Register");
SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
em_sysctl_reg_handler, "IU",
"Receiver Control Register");
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
"Flow Control High Watermark");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
"Flow Control Low Watermark");
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct tx_ring *txr = &tx_que->txr;
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
CTLFLAG_RD, NULL, "TX Queue Name");
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
queue_list = SYSCTL_CHILDREN(queue_node);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, adapter,
E1000_TDH(txr->me),
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
em_sysctl_reg_handler, "IU",
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
"Transmit Descriptor Head");
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, adapter,
E1000_TDT(txr->me),
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
em_sysctl_reg_handler, "IU",
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
"Transmit Descriptor Tail");
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
CTLFLAG_RD, &txr->tx_irq,
"Queue MSI-X Transmit Interrupts");
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct rx_ring *rxr = &rx_que->rxr;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
CTLFLAG_RD, NULL, "RX Queue Name");
queue_list = SYSCTL_CHILDREN(queue_node);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, adapter,
E1000_RDH(rxr->me),
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
em_sysctl_reg_handler, "IU",
"Receive Descriptor Head");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
Specify a CTLTYPE_FOO so that a future sysctl(8) change does not need to rely on the format string.
2011-01-18 21:14:23 +00:00
CTLTYPE_UINT | CTLFLAG_RD, adapter,
E1000_RDT(rxr->me),
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
em_sysctl_reg_handler, "IU",
"Receive Descriptor Tail");
SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
CTLFLAG_RD, &rxr->rx_irq,
"Queue MSI-X Receive Interrupts");
}
/* MAC stats get their own sub node */
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, NULL, "Statistics");
stat_list = SYSCTL_CHILDREN(stat_node);
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &stats->ecol,
"Excessive collisions");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
CTLFLAG_RD, &stats->scc,
"Single collisions");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
CTLFLAG_RD, &stats->mcc,
"Multiple collisions");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
CTLFLAG_RD, &stats->latecol,
"Late collisions");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
CTLFLAG_RD, &stats->colc,
"Collision Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.symerrs,
"Symbol Errors");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.sec,
"Sequence Errors");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.dc,
"Defer Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.mpc,
"Missed Packets");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.rnbc,
"Receive No Buffers");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
CTLFLAG_RD, &adapter->stats.ruc,
"Receive Undersize");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
CTLFLAG_RD, &adapter->stats.rfc,
"Fragmented Packets Received ");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
CTLFLAG_RD, &adapter->stats.roc,
"Oversized Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
Tweak the stats exported by the e1000 drivers: - Add a single sysctl procedure to all three drivers to read an arbitrary register (the register is passed as arg2). Use it to replace existing routines in igb(4) that used a separate routine for each register, and to add support for missing stats in em(4) and lem(4). - Move the 'rx_overruns' and 'watchdog_timeouts' stats out of the MAC stats section as they are driver stats, not MAC counters. - Simplify the code that creates per-queue stats in igb(4) to use a single loop and remove duplicated code. - Properly read all 64 bits of the 'good octets received/transmitted' in em(4) and lem(4). - Actually read the interrupt count registers in em(4), and drop the 'host to card' sysctl stats from em(4) as they are not implemented in any of the hardware this driver supports. - Restore several stats to em(4) that were lost in the earlier stats conversion including per-queue stats. - Export several MAC stats in em(4) that were exported in igb(4) but not in em(4). - Export stats in lem(4) using individual sysctls as in em(4) and igb(4). Reviewed by: jfv MFC after: 1 week
2010-09-20 16:04:44 +00:00
CTLFLAG_RD, &adapter->stats.rjc,
"Recevied Jabber");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.rxerrc,
"Receive Errors");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.crcerrs,
"CRC errors");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.algnerrc,
"Alignment Errors");
/* On 82575 these are collision counts */
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.cexterr,
"Collision/Carrier extension errors");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.xonrxc,
"XON Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.xontxc,
"XON Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.xoffrxc,
"XOFF Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.xofftxc,
"XOFF Transmitted");
/* Packet Reception Stats */
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.tpr,
"Total Packets Received ");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.gprc,
"Good Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.bprc,
"Broadcast Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.mprc,
"Multicast Packets Received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.prc64,
"64 byte frames received ");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.prc127,
"65-127 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.prc255,
"128-255 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.prc511,
"256-511 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.prc1023,
"512-1023 byte frames received");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.prc1522,
"1023-1522 byte frames received");
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
CTLFLAG_RD, &adapter->stats.gorc,
"Good Octets Received");
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
/* Packet Transmission Stats */
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
CTLFLAG_RD, &adapter->stats.gotc,
"Good Octets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.tpt,
"Total Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.gptc,
"Good Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.bptc,
"Broadcast Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.mptc,
"Multicast Packets Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ptc64,
"64 byte frames transmitted ");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ptc127,
"65-127 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ptc255,
"128-255 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ptc511,
"256-511 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ptc1023,
"512-1023 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ptc1522,
"1024-1522 byte frames transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.tsctc,
"TSO Contexts Transmitted");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.tsctfc,
"TSO Contexts Failed");
/* Interrupt Stats */
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, NULL, "Interrupt Statistics");
int_list = SYSCTL_CHILDREN(int_node);
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.iac,
"Interrupt Assertion Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.icrxptc,
"Interrupt Cause Rx Pkt Timer Expire Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.icrxatc,
"Interrupt Cause Rx Abs Timer Expire Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ictxptc,
"Interrupt Cause Tx Pkt Timer Expire Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ictxatc,
"Interrupt Cause Tx Abs Timer Expire Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ictxqec,
"Interrupt Cause Tx Queue Empty Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.ictxqmtc,
"Interrupt Cause Tx Queue Min Thresh Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.icrxdmtc,
"Interrupt Cause Rx Desc Min Thresh Count");
sysctl(9) cleanup checkpoint: amd64 GENERIC builds cleanly. Commit the Intel drivers.
2011-01-12 19:53:23 +00:00
SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
CTLFLAG_RD, &adapter->stats.icrxoc,
"Interrupt Cause Receiver Overrun Count");
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
/**********************************************************************
*
* This routine provides a way to dump out the adapter eeprom,
* often a useful debug/service tool. This only dumps the first
* 32 words, stuff that matters is in that extent.
*
**********************************************************************/
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
static int
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
{
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
struct adapter *adapter = (struct adapter *)arg1;
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
int error;
int result;
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
result = -1;
error = sysctl_handle_int(oidp, &result, 0, req);
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
if (error || !req->newptr)
return (error);
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
/*
* This value will cause a hex dump of the
* first 32 16-bit words of the EEPROM to
* the screen.
*/
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
if (result == 1)
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
em_print_nvm_info(adapter);
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (error);
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
}
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
static void
em_print_nvm_info(struct adapter *adapter)
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
u16 eeprom_data;
int i, j, row = 0;
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
/* Its a bit crude, but it gets the job done */
printf("\nInterface EEPROM Dump:\n");
printf("Offset\n0x0000 ");
for (i = 0, j = 0; i < 32; i++, j++) {
if (j == 8) { /* Make the offset block */
j = 0; ++row;
printf("\n0x00%x0 ",row);
}
e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
printf("%04x ", eeprom_data);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
}
Move statistics into the sysctl tree making it easier to find and use them. Add previously hidden statistics, some of which include interrupt and host/card communication counters.
2010-06-16 20:57:41 +00:00
printf("\n");
Add support for Quad port adapter Add sysctl's to display statistics/debug_info Set WAIT_FOR_AUTONEG_DEFAULT to zero by default Increment packet in/out statistics inline instead of every two seconds. MFC after: 3 days
2003-06-05 17:51:38 +00:00
}
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
static int
em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
{
struct em_int_delay_info *info;
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
struct adapter *adapter;
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
u32 regval;
Update to igb and em: em revision 7.0.0: - Using driver devclass, seperate legacy (pre-pcie) code into a seperate source file. This will at least help protect against regression issues. It compiles along with em, and is transparent to end use, devices in each appear to be 'emX'. When using em in a modular form this also allows the legacy stuff to be defined out. - Add tx and rx rings as in igb, in the 82574 this becomes actual multiqueue for the first time (2 queues) while in other PCIE adapters its just make code cleaner. - Add RX mbuf handling logic that matches igb, this will eliminate packet drops due to temporary mbuf shortage. igb revision 1.9.3: - Following the ixgbe code, use a new approach in what was called 'get_buf', the routine now has been made independent of rxeof, it now does the update to the engine TDT register, this design allows temporary mbuf resources to become non-critical, not requiring a packet to be discarded, instead it just returns and does not increment the tail pointer. - With the above change it was also unnecessary to keep 'spare' maps around, since we do not have the discard issue. - Performance tweaks and improvements to the code also. MFC in a week
2010-03-29 23:36:34 +00:00
int error, usecs, ticks;
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
info = (struct em_int_delay_info *) arg1;
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
usecs = info->value;
error = sysctl_handle_int(oidp, &usecs, 0, req);
if (error != 0 || req->newptr == NULL)
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (error);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (EINVAL);
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
info->value = usecs;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
ticks = EM_USECS_TO_TICKS(usecs);
if_lem.c: make sure that lem_rxeof() can drain the entire rx queue irrespective of the setting of lem_rx_process_limit, while giving a chance to the taskqueue scheduler to act after each chunk. This makes lem_rxeof similar to the one in if_em.c and if_igb.c . if_lem.c and if_em.c: add a sysctl to manually configure the 'itr' moderation register. Approved by: Jack Vogel
2013-05-09 17:07:30 +00:00
if (info->offset == E1000_ITR) /* units are 256ns here */
ticks *= 4;
Some more minor cleanups of em(4) driver: - Destroy mutex in case of attach failure. [1] - Lock properly em_watchdog(). [1] - Lock properly em_sysctl_int_delay(). [1] - Remove unused global adapter linked list. - Remove unused dma_size field from struct em_dma_alloc. - Do not touch interface statistics, that must be edited only by upper layers. [1] Submitted by: yongari [1]
2005-10-20 09:55:49 +00:00
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
adapter = info->adapter;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
regval = (regval & ~0xffff) | (ticks & 0xffff);
/* Handle a few special cases. */
switch (info->offset) {
case E1000_RDTR:
break;
case E1000_TIDV:
if (ticks == 0) {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
/* Don't write 0 into the TIDV register. */
regval++;
} else
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
adapter->txd_cmd |= E1000_TXD_CMD_IDE;
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
break;
}
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
return (0);
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
}
static void
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
const char *description, struct em_int_delay_info *info,
int offset, int value)
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
{
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
info->adapter = adapter;
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
info->offset = offset;
info->value = value;
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
Add facilities for tuning the "em" driver's interrupt delays without recompiling the driver. See the comments near the top of "if_em.h" for descriptions of these delays. Four new loader tunables control the system-wide default values: hw.em.tx_int_delay hw.em.rx_int_delay hw.em.tx_abs_int_delay hw.em.rx_abs_int_delay The tunables are specified in microseconds. The valid range is 0-67108 usec., and 0 means that the timer is disabled. There are also four new sysctls (actually, a set of four for each "em" device in the system) to query and change the interrupt delays after the system is up: hw.em0.tx_int_delay hw.em0.rx_int_delay hw.em0.tx_abs_int_delay (not present for 82542/3/4 adapters) hw.em0.rx_abs_int_delay (not present for 82542/3/4 adapters) It seems to be OK to change these values even while the adapter is passing traffic. Approved by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> MFC after: 4 weeks
2003-08-01 17:33:59 +00:00
OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
info, 0, em_sysctl_int_delay, "I", description);
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* Set flow control using sysctl:
* Flow control values:
* 0 - off
* 1 - rx pause
* 2 - tx pause
* 3 - full
*/
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
static int
em_set_flowcntl(SYSCTL_HANDLER_ARGS)
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
{
int error;
static int input = 3; /* default is full */
struct adapter *adapter = (struct adapter *) arg1;
error = sysctl_handle_int(oidp, &input, 0, req);
if ((error) || (req->newptr == NULL))
return (error);
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
if (input == adapter->fc) /* no change? */
return (error);
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
switch (input) {
case e1000_fc_rx_pause:
case e1000_fc_tx_pause:
case e1000_fc_full:
case e1000_fc_none:
adapter->hw.fc.requested_mode = input;
adapter->fc = input;
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
break;
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
default:
/* Do nothing */
return (error);
}
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
e1000_force_mac_fc(&adapter->hw);
return (error);
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
}
Change the interface to the Energy Efficient Ethernet (EEE) setting in the igb and em driver. This was necessitated by a shared code change that I was given late in the game, a data type changed from bool to int, in the last update I dealt with it by a cast, but it was pointed out (thanks jhb) that there was a potential problem with this. John suggested this safer approach, and it is fine with me... MFC after:2 days (to catch the 9.1 update)
2012-07-07 20:21:05 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* Manage Energy Efficient Ethernet:
* Control values:
* 0/1 - enabled/disabled
*/
Change the interface to the Energy Efficient Ethernet (EEE) setting in the igb and em driver. This was necessitated by a shared code change that I was given late in the game, a data type changed from bool to int, in the last update I dealt with it by a cast, but it was pointed out (thanks jhb) that there was a potential problem with this. John suggested this safer approach, and it is fine with me... MFC after:2 days (to catch the 9.1 update)
2012-07-07 20:21:05 +00:00
static int
em_sysctl_eee(SYSCTL_HANDLER_ARGS)
{
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
struct adapter *adapter = (struct adapter *) arg1;
int error, value;
Change the interface to the Energy Efficient Ethernet (EEE) setting in the igb and em driver. This was necessitated by a shared code change that I was given late in the game, a data type changed from bool to int, in the last update I dealt with it by a cast, but it was pointed out (thanks jhb) that there was a potential problem with this. John suggested this safer approach, and it is fine with me... MFC after:2 days (to catch the 9.1 update)
2012-07-07 20:21:05 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
value = adapter->hw.dev_spec.ich8lan.eee_disable;
error = sysctl_handle_int(oidp, &value, 0, req);
if (error || req->newptr == NULL)
return (error);
adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
em_if_init(adapter->ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
return (0);
Change the interface to the Energy Efficient Ethernet (EEE) setting in the igb and em driver. This was necessitated by a shared code change that I was given late in the game, a data type changed from bool to int, in the last update I dealt with it by a cast, but it was pointed out (thanks jhb) that there was a potential problem with this. John suggested this safer approach, and it is fine with me... MFC after:2 days (to catch the 9.1 update)
2012-07-07 20:21:05 +00:00
}
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
static int
em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter;
int error;
int result;
This delta has a few important items: PR 122839 is fixed in both em and in igb Second, the issue on building modules since the static kernel build changes is now resolved. I was not able to get the fancier directory hierarchy working, but this works, both em and igb build as modules now. Third, there is now support in em for two new NICs, Hartwell (or 82574) is a low cost PCIE dual port adapter that has MSIX, for this release it uses 3 vectors only, RX, TX, and LINK. In the next release I will add a second TX and RX queue. Also, there is support here for ICH10, the followon to ICH9. Both of these are early releases, general availability will follow soon. Fourth: On Hartwell and ICH10 we now have IEEE 1588 PTP support, I have implemented this in a provisional way so that early adopters may try and comment on the functionality. The IOCTL structure may change. This feature is off by default, you need to edit the Makefile and add the EM_TIMESYNC define to get the code. Enjoy all!!
2008-04-25 21:19:41 +00:00
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
result = -1;
error = sysctl_handle_int(oidp, &result, 0, req);
if (error || !req->newptr)
return (error);
if (result == 1) {
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
adapter = (struct adapter *) arg1;
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
em_print_debug_info(adapter);
Some trivial style(9) fixes - Delete trailing whitespace. - Fix leading indentation (convert single column spaces to tabs). - Convert "[Ff]all through" to "FALLTHROUGH", per implicit project style/spelling. Reviewed by: sbruno Differential Revision: D11665
2017-07-20 04:32:06 +00:00
}
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
return (error);
}
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
static int
em_get_rs(SYSCTL_HANDLER_ARGS)
{
struct adapter *adapter = (struct adapter *) arg1;
int error;
int result;
result = 0;
error = sysctl_handle_int(oidp, &result, 0, req);
if (error || !req->newptr || result != 1)
return (error);
em_dump_rs(adapter);
return (error);
}
static void
em_if_debug(if_ctx_t ctx)
{
em_dump_rs(iflib_get_softc(ctx));
}
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
/*
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
* This routine is meant to be fluid, add whatever is
* needed for debugging a problem. -jfv
*/
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
static void
em_print_debug_info(struct adapter *adapter)
{
Access *correct* ifp data structure when debug sysctl is invoked. Submitted by: Kevin Bowling <kevin.bowling@kev009.com> Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D10126
2017-03-27 15:08:02 +00:00
device_t dev = iflib_get_dev(adapter->ctx);
struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct tx_ring *txr = &adapter->tx_queues->txr;
struct rx_ring *rxr = &adapter->rx_queues->rxr;
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
Access *correct* ifp data structure when debug sysctl is invoked. Submitted by: Kevin Bowling <kevin.bowling@kev009.com> Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D10126
2017-03-27 15:08:02 +00:00
if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
printf("Interface is RUNNING ");
else
printf("Interface is NOT RUNNING\n");
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
Access *correct* ifp data structure when debug sysctl is invoked. Submitted by: Kevin Bowling <kevin.bowling@kev009.com> Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D10126
2017-03-27 15:08:02 +00:00
if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
printf("and INACTIVE\n");
Part 2 of 2 New deltas for the 1G drivers. There have still been intermittent problems with apparent TX hangs for some customers. These have been problematic to reproduce but I believe these changes will address them. Testing on a number of fronts have been positive. EM: there is an important 'chicken bit' fix for 82574 in the shared code this is supported in the core here. - The TX path has been tightened up to improve performance. In particular UDP with jumbo frames was having problems, and the changes here have improved that. - OACTIVE has been used more carefully on the theory that some hangs may be due to a problem in this interaction - Problems with the RX init code, the "lazy" allocation and ring initialization has been found to cause problems in some newer client systems, and as it really is not that big a win (its not in a hot path) it seems best to remove it. - HWTSO was broken when VLAN HWTAGGING or HWFILTER is used, I found this was due to an error in setting up the descriptors in em_xmit. IGB: - TX is also improved here. With multiqueue I realized its very important to handle OACTIVE only under the CORE lock so there are no races between the queues. - Flow Control handling was broken in a couple ways, I have changed and I hope improved that in this delta. - UDP also had a problem in the TX path here, it was change to improve that. - On some hardware, with the driver static, a weird stray interrupt seems to sometimes fire and cause a panic in the RX mbuf refresh code. This is addressed by setting interrupts late in the init path, and also to set all interrupts bits off at the start of that.
2011-12-10 07:08:52 +00:00
else
printf("and ACTIVE\n");
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
device_printf(dev, "TX Queue %d ------\n", i);
device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
}
for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
device_printf(dev, "RX Queue %d ------\n", j);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
IFLIB updates - unconditionally enable BUS_DMA on non-x86 architectures - speed up rxd zeroing via customized function - support out of order updates to rxd's - add prefetching to hardware descriptor rings - only prefetch on 10G or faster hardware - add seperate tx queue intr function - preliminary rework of NETMAP interfaces, WIP Submitted by: Matt Macy <mmacy@nextbsd.org> Sponsored by: Limelight Networks
2017-03-13 22:53:06 +00:00
E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
}
Bug fix delta to the em driver: - Chasin down bogus watchdogs has led to an improved design to this handling, the hang decision takes place in the tx cleanup, with only a simple report check in local_timer. Our tests have shown no false watchdogs with this code. - VLAN fixes from jhb, the shadow vfta should be per interface, but as global it was not. Thanks John. - Bug fixes in the support for new PCH2 hardware. - Thanks for all the help and feedback on the driver, changes to lem with be coming shortly as well.
2010-10-26 00:07:58 +00:00
}
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
/*
* 82574 only:
* Write a new value to the EEPROM increasing the number of MSIX
* vectors from 3 to 5, for proper multiqueue support.
*/
static void
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
em_enable_vectors_82574(if_ctx_t ctx)
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
{
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
struct adapter *adapter = iflib_get_softc(ctx);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
struct e1000_hw *hw = &adapter->hw;
Migrate e1000 to the IFLIB framework: - em(4) igb(4) and lem(4) - deprecate the igb device from kernel configurations - create a symbolic link in /boot/kernel from if_em.ko to if_igb.ko Devices tested: - 82574L - I218-LM - 82546GB - 82579LM - I350 - I217 Please report problems to freebsd-net@freebsd.org Partial review from jhb and suggestions on how to *not* brick folks who originally would have lost their igbX device. Submitted by: mmacy@nextbsd.org MFC after: 2 weeks Relnotes: yes Sponsored by: Limelight Networks and Dell EMC Isilon Differential Revision: https://reviews.freebsd.org/D8299
2017-01-10 03:23:22 +00:00
device_t dev = iflib_get_dev(ctx);
Change EM_MULTIQUEUE to a real kernconf entry and enable support for up to 2 rx/tx queues for the 82574. Program the 82574 to enable 5 msix vectors, assign 1 to each rx queue, 1 to each tx queue and 1 to the link handler. Inspired by DragonFlyBSD, enable some RSS logic for handling tx queue handling/processing. Move multiqueue handler functions so that they line up better in a diff review to if_igb.c Always enqueue tx work to be done in em_mq_start, if unable to acquire the TX lock, then this will be processed in the background later by the taskqueue. Remove mbuf argument from em_start_mq_locked() as the work is always enqueued. (stolen from igb) Setup TARC, TXDCTL and RXDCTL registers for better performance and stability in multiqueue and singlequeue implementations. Handle Intel errata 3 and generic multiqueue behavior with the initialization of TARC(0) and TARC(1) Bind interrupt threads to cpus in order. (stolen from igb) Add 2 new DDB functions, one to display the queue(s) and their settings and one to reset the adapter. Primarily used for debugging. In the multiqueue configuration, bump RXD and TXD ring size to max for the adapter (4096). Setup an RDTR of 64 and an RADV of 128 in multiqueue configuration to cut down on the number of interrupts. RADV was arbitrarily set to 2x RDTR and can be adjusted as needed. Cleanup the display in top a bit to make it clearer where the taskqueue threads are running and what they should be doing. Ensure that both queues are processed by em_local_timer() by writing them both to the IMS register to generate soft interrupts. Ensure that an soft interrupt is generated when em_msix_link() is run so that any races between assertion of the link/status interrupt and a rx/tx interrupt are handled. Document existing tuneables: hw.em.eee_setting, hw.em.msix, hw.em.smart_pwr_down, hw.em.sbp Document use of hw.em.num_queues and the new kernel option EM_MULTIQUEUE Thanks to Intel for their continued support of FreeBSD. Reviewed by: erj jfv hiren gnn wblock Obtained from: Intel Corporation MFC after: 2 weeks Relnotes: Yes Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D1994
2015-06-03 18:01:09 +00:00
u16 edata;
e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
printf("Current cap: %#06x\n", edata);
if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
device_printf(dev, "Writing to eeprom: increasing "
"reported MSIX vectors from 3 to 5...\n");
edata &= ~(EM_NVM_MSIX_N_MASK);
edata |= 4 << EM_NVM_MSIX_N_SHIFT;
e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
e1000_update_nvm_checksum(hw);
device_printf(dev, "Writing to eeprom: done\n");
}
}
Reference in New Issue Copy Permalink