d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
73fa18d121
freebsd-skq/sys/dev/e1000/if_em.c

5125 lines
144 KiB
C
Raw Normal View History

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
/******************************************************************************
Update the 1G drivers, shared code sync with Intel, igb now has a queue notion that has a single interrupt with an RX/TX pair, this will reduce the total interrupts seen on a system. Both em and igb have a new watchdog method. igb has fixes from Pyun Yong-Hyeon that have improved stability, thank you :) I wish to MFC this for 7.3 asap, please test if able.
2010-01-26 22:32:22 +00:00
Copyright (c) 2001-2010, Intel Corporation
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
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
3. Neither the name of the Intel Corporation nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS 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 COPYRIGHT OWNER 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.
******************************************************************************/
/*$FreeBSD$*/
- Added support for receive in multiple descriptors. This simplifies code for jumbo frames. - Cleaned up coding conventions to make code more unix-like. - Cleaned up code in if_em_fxhw.c and if_em_phy.c. Added relevant comments. MFC after: 1 week
2002-02-13 18:19:27 +00:00
- Don't pollute opt_global.h with DEVICE_POLLING and introduce opt_device_polling.h - Include opt_device_polling.h into appropriate files. - Embrace with HAVE_KERNEL_OPTION_HEADERS the include in the files that can be compiled as loadable modules. Reviewed by: bde
2005-10-05 10:09:17 +00:00
#ifdef HAVE_KERNEL_OPTION_HEADERS
#include "opt_device_polling.h"
Restore opt_inet.h include which was lost in the last commit.
2008-12-07 21:32:56 +00:00
#include "opt_inet.h"
- Don't pollute opt_global.h with DEVICE_POLLING and introduce opt_device_polling.h - Include opt_device_polling.h into appropriate files. - Embrace with HAVE_KERNEL_OPTION_HEADERS the include in the files that can be compiled as loadable modules. Reviewed by: bde
2005-10-05 10:09:17 +00:00
#endif
Move includes from if_em.h to if_em.c and sort them.
2006-02-14 13:11:36 +00:00
#include <sys/param.h>
#include <sys/systm.h>
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 __FreeBSD_version >= 800000
- 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
#include <sys/buf_ring.h>
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
#endif
Move includes from if_em.h to if_em.c and sort them.
2006-02-14 13:11:36 +00:00
#include <sys/bus.h>
#include <sys/endian.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/module.h>
#include <sys/rman.h>
#include <sys/socket.h>
#include <sys/sockio.h>
#include <sys/sysctl.h>
#include <sys/taskqueue.h>
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
#include <sys/eventhandler.h>
Move includes from if_em.h to if_em.c and sort them.
2006-02-14 13:11:36 +00:00
#include <machine/bus.h>
#include <machine/resource.h>
#include <net/bpf.h>
#include <net/ethernet.h>
#include <net/if.h>
#include <net/if_arp.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/if_types.h>
#include <net/if_vlan_var.h>
#include <netinet/in_systm.h>
#include <netinet/in.h>
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +00:00
#include <netinet/if_ether.h>
Move includes from if_em.h to if_em.c and sort them.
2006-02-14 13:11:36 +00:00
#include <netinet/ip.h>
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
#include <netinet/ip6.h>
Move includes from if_em.h to if_em.c and sort them.
2006-02-14 13:11:36 +00:00
#include <netinet/tcp.h>
#include <netinet/udp.h>
Second attempt at fixing module build Pointyhat: pdeuskar
2006-09-09 20:05:24 +00:00
#include <machine/in_cksum.h>
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
#include <dev/led/led.h>
Move includes from if_em.h to if_em.c and sort them.
2006-02-14 13:11:36 +00:00
#include <dev/pci/pcivar.h>
#include <dev/pci/pcireg.h>
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 "e1000_api.h"
Thanks to report from Neil Hoggarth I found a missing UNLOCK in the watchdog code. This delta also incorporates some missing PCI IDs that got added. PR 122928 - might be fixed by this, no verification at this point.
2008-05-21 21:06:17 +00:00
#include "e1000_82571.h"
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"
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
* Set this to one to display debug statistics
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
*********************************************************************/
- 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 em_display_debug_stats = 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
/*********************************************************************
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
*********************************************************************/
Add a missing fragment in the tx msix handler to invoke another if all work is not done. Sync the igb driver with changes suggested by yongari and made in em, these made sense to be in both drivers.
2010-04-14 20:55:33 +00:00
char em_driver_version[] = "7.0.5";
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
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 em_vendor_info_t em_vendor_info_array[] =
{
- 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
/* Intel(R) PRO/1000 Network Connection */
{ 0x8086, E1000_DEV_ID_82571EB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82571EB_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82571EB_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0},
Thanks to report from Neil Hoggarth I found a missing UNLOCK in the watchdog code. This delta also incorporates some missing PCI IDs that got added. PR 122928 - might be fixed by this, no verification at this point.
2008-05-21 21:06:17 +00:00
{ 0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL,
PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD,
PCI_ANY_ID, PCI_ANY_ID, 0},
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
{ 0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER,
PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP,
PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER,
Merge new vendor release - 6.2.9. Details: o if_em.c changes: - Added several new PCI ids. - Check em_check_phy_reset_block() before doing SIOCSIFMEDIA ioctl. - Don't touch TARC registers, they are now handled in shared code in if_em_hw.c. - Move RDH and RDT setting to the end of em_initialize_receive_unit(). - Declare em_read_pcie_cap_reg(), now empty. o if_em_hw.c dropped in from vendor, then restored rev. 1.15. o if_em_hw.h dropped in from vendor, then modified: - Added RX overrun interrupt flag to interrupt enable mask. - Remove declarations of em_io_read(), em_io_write(). Approved by: jfv
2006-10-31 15:00:14 +00:00
PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER,
PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_82572EI_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82572EI_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82572EI_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0},
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
{ 0x8086, E1000_DEV_ID_82572EI, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_82573E, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82573E_IAMT, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82573L, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_82583V, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT,
PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT,
PCI_ANY_ID, PCI_ANY_ID, 0},
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
{ 0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT,
PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT,
PCI_ANY_ID, PCI_ANY_ID, 0},
Merge new vendor release - 6.2.9. Details: o if_em.c changes: - Added several new PCI ids. - Check em_check_phy_reset_block() before doing SIOCSIFMEDIA ioctl. - Don't touch TARC registers, they are now handled in shared code in if_em_hw.c. - Move RDH and RDT setting to the end of em_initialize_receive_unit(). - Declare em_read_pcie_cap_reg(), now empty. o if_em_hw.c dropped in from vendor, then restored rev. 1.15. o if_em_hw.h dropped in from vendor, then modified: - Added RX overrun interrupt flag to interrupt enable mask. - Remove declarations of em_io_read(), em_io_write(). Approved by: jfv
2006-10-31 15:00:14 +00:00
{ 0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_ICH8_IGP_AMT, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH8_IGP_C, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH8_IFE, PCI_ANY_ID, PCI_ANY_ID, 0},
Merge new vendor release - 6.2.9. Details: o if_em.c changes: - Added several new PCI ids. - Check em_check_phy_reset_block() before doing SIOCSIFMEDIA ioctl. - Don't touch TARC registers, they are now handled in shared code in if_em_hw.c. - Move RDH and RDT setting to the end of em_initialize_receive_unit(). - Declare em_read_pcie_cap_reg(), now empty. o if_em_hw.c dropped in from vendor, then restored rev. 1.15. o if_em_hw.h dropped in from vendor, then modified: - Added RX overrun interrupt flag to interrupt enable mask. - Remove declarations of em_io_read(), em_io_write(). Approved by: jfv
2006-10-31 15:00:14 +00:00
{ 0x8086, E1000_DEV_ID_ICH8_IFE_GT, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH8_IFE_G, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH8_IGP_M, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_ICH8_82567V_3, PCI_ANY_ID, PCI_ANY_ID, 0},
Thanks to report from Neil Hoggarth I found a missing UNLOCK in the watchdog code. This delta also incorporates some missing PCI IDs that got added. PR 122928 - might be fixed by this, no verification at this point.
2008-05-21 21:06:17 +00:00
{ 0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_ICH9_IGP_AMT, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH9_IGP_C, PCI_ANY_ID, PCI_ANY_ID, 0},
Thanks to report from Neil Hoggarth I found a missing UNLOCK in the watchdog code. This delta also incorporates some missing PCI IDs that got added. PR 122928 - might be fixed by this, no verification at this point.
2008-05-21 21:06:17 +00:00
{ 0x8086, E1000_DEV_ID_ICH9_IGP_M, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH9_IGP_M_V, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_ICH9_IFE, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH9_IFE_GT, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH9_IFE_G, PCI_ANY_ID, PCI_ANY_ID, 0},
This small change will allow this driver in HEAD to build on 6.3 as well as 7 :)
2008-05-19 23:01:11 +00:00
{ 0x8086, E1000_DEV_ID_ICH9_BM, PCI_ANY_ID, PCI_ANY_ID, 0},
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
{ 0x8086, E1000_DEV_ID_82574L, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_82574LA, PCI_ANY_ID, PCI_ANY_ID, 0},
Thanks to report from Neil Hoggarth I found a missing UNLOCK in the watchdog code. This delta also incorporates some missing PCI IDs that got added. PR 122928 - might be fixed by this, no verification at this point.
2008-05-21 21:06:17 +00:00
{ 0x8086, E1000_DEV_ID_ICH10_R_BM_LM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH10_R_BM_LF, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH10_R_BM_V, PCI_ANY_ID, PCI_ANY_ID, 0},
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
{ 0x8086, E1000_DEV_ID_ICH10_D_BM_LM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_ICH10_D_BM_LF, PCI_ANY_ID, PCI_ANY_ID, 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
{ 0x8086, E1000_DEV_ID_PCH_M_HV_LM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_PCH_M_HV_LC, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_PCH_D_HV_DM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_PCH_D_HV_DC, PCI_ANY_ID, PCI_ANY_ID, 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
/* required last entry */
{ 0, 0, 0, 0, 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
};
/*********************************************************************
* Table of branding strings for all supported NICs.
*********************************************************************/
static char *em_strings[] = {
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
"Intel(R) PRO/1000 Network Connection"
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
*********************************************************************/
- 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_probe(device_t);
static int em_attach(device_t);
static int em_detach(device_t);
static int em_shutdown(device_t);
static int em_suspend(device_t);
static int em_resume(device_t);
static void em_start(struct ifnet *);
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 em_start_locked(struct ifnet *, struct tx_ring *);
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
#ifdef EM_MULTIQUEUE
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 int em_mq_start(struct ifnet *, struct mbuf *);
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_mq_start_locked(struct ifnet *,
struct tx_ring *, struct mbuf *);
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_qflush(struct ifnet *);
#endif
- 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_ioctl(struct ifnet *, u_long, caddr_t);
static void em_init(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
static void em_init_locked(struct 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
static void em_stop(void *);
static void em_media_status(struct ifnet *, struct ifmediareq *);
static int em_media_change(struct ifnet *);
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_identify_hardware(struct adapter *);
static int em_allocate_pci_resources(struct 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
static int em_allocate_legacy(struct adapter *);
static int em_allocate_msix(struct adapter *);
static int em_allocate_queues(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_setup_msix(struct adapter *);
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_free_pci_resources(struct 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
static void em_local_timer(void *);
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 em_reset(struct adapter *);
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_setup_interface(device_t, struct 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
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_setup_transmit_structures(struct adapter *);
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_initialize_transmit_unit(struct 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
static int em_allocate_transmit_buffers(struct tx_ring *);
static void em_free_transmit_structures(struct adapter *);
static void em_free_transmit_buffers(struct tx_ring *);
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 int em_setup_receive_structures(struct 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
static int em_allocate_receive_buffers(struct rx_ring *);
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_initialize_receive_unit(struct 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
static void em_free_receive_structures(struct adapter *);
static void em_free_receive_buffers(struct rx_ring *);
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_enable_intr(struct adapter *);
static void em_disable_intr(struct adapter *);
static void em_update_stats_counters(struct 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
static bool em_txeof(struct tx_ring *);
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
static bool em_rxeof(struct rx_ring *, int, int *);
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
#ifndef __NO_STRICT_ALIGNMENT
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_fixup_rx(struct rx_ring *);
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
#endif
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 em_receive_checksum(struct e1000_rx_desc *, struct mbuf *);
static void em_transmit_checksum_setup(struct tx_ring *, struct mbuf *,
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 *, u32 *);
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 bool em_tso_setup(struct tx_ring *, struct mbuf *, u32 *, u32 *);
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_set_promisc(struct adapter *);
static void em_disable_promisc(struct adapter *);
static void em_set_multi(struct adapter *);
static void em_print_hw_stats(struct adapter *);
static void em_update_link_status(struct 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
static void em_refresh_mbufs(struct rx_ring *, int);
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 em_register_vlan(void *, struct ifnet *, u16);
static void em_unregister_vlan(void *, struct ifnet *, u16);
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 *);
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_xmit(struct tx_ring *, struct mbuf **);
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 int em_dma_malloc(struct adapter *, bus_size_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
struct em_dma_alloc *, int);
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_dma_free(struct adapter *, struct em_dma_alloc *);
static void em_print_debug_info(struct adapter *);
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 *);
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_stats(SYSCTL_HANDLER_ARGS);
static int em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
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 *);
static void em_get_hw_control(struct adapter *);
static void em_release_hw_control(struct 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
static void em_get_wakeup(device_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
static void em_enable_wakeup(device_t);
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 *);
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 em_led_func(void *, int);
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
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_irq_fast(void *);
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 */
static void em_msix_tx(void *);
static void em_msix_rx(void *);
static void em_msix_link(void *);
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 void em_handle_tx(void *context, int pending);
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 em_handle_rx(void *context, int pending);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
static void em_handle_link(void *context, int pending);
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
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
static void em_add_rx_process_limit(struct adapter *, const char *,
const char *, int *, int);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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
#ifdef DEVICE_POLLING
static poll_handler_t em_poll;
#endif /* POLLING */
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 */
DEVMETHOD(device_probe, em_probe),
DEVMETHOD(device_attach, em_attach),
DEVMETHOD(device_detach, em_detach),
DEVMETHOD(device_shutdown, em_shutdown),
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
DEVMETHOD(device_suspend, em_suspend),
DEVMETHOD(device_resume, em_resume),
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
{0, 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
};
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
};
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
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);
MODULE_DEPEND(em, pci, 1, 1, 1);
MODULE_DEPEND(em, ether, 1, 1, 1);
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)
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
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
/* Allow common code without TSO */
#ifndef CSUM_TSO
#define CSUM_TSO 0
#endif
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 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_INT("hw.em.tx_int_delay", &em_tx_int_delay_dflt);
TUNABLE_INT("hw.em.rx_int_delay", &em_rx_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
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 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_INT("hw.em.tx_abs_int_delay", &em_tx_abs_int_delay_dflt);
TUNABLE_INT("hw.em.rx_abs_int_delay", &em_rx_abs_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
static int em_rxd = EM_DEFAULT_RXD;
static int em_txd = EM_DEFAULT_TXD;
- 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
TUNABLE_INT("hw.em.rxd", &em_rxd);
TUNABLE_INT("hw.em.txd", &em_txd);
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;
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
TUNABLE_INT("hw.em.smart_pwr_down", &em_smart_pwr_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
/* Controls whether promiscuous also shows bad packets */
static int em_debug_sbp = FALSE;
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
TUNABLE_INT("hw.em.sbp", &em_debug_sbp);
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
/* Local controls for MSI/MSIX */
Address the LOD that some are seeing, put the RX lock back in rxeof (I could see little point in taking it out), and now release it before the stack entry. Also, make it so the 82574 does not configure for multiqueue when its not used in the stack.
2010-04-28 19:22:52 +00:00
#ifdef EM_MULTIQUEUE
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_enable_msix = TRUE;
static int em_msix_queues = 2; /* for 82574, can be 1 or 2 */
Address the LOD that some are seeing, put the RX lock back in rxeof (I could see little point in taking it out), and now release it before the stack entry. Also, make it so the 82574 does not configure for multiqueue when its not used in the stack.
2010-04-28 19:22:52 +00:00
#else
static int em_enable_msix = FALSE;
static int em_msix_queues = 0; /* disable */
#endif
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
TUNABLE_INT("hw.em.enable_msix", &em_enable_msix);
TUNABLE_INT("hw.em.msix_queues", &em_msix_queues);
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;
TUNABLE_INT("hw.em.rx_process_limit", &em_rx_process_limit);
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
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
/* Flow control setting - default to FULL */
static int em_fc_setting = e1000_fc_full;
TUNABLE_INT("hw.em.fc_setting", &em_fc_setting);
/*
** Shadow VFTA table, this is needed because
** the real vlan filter table gets cleared during
** a soft reset and the driver needs to be able
** to repopulate it.
*/
static u32 em_shadow_vfta[EM_VFTA_SIZE];
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
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 identification routine
*
* em_probe determines if the driver should be loaded on
* adapter based on PCI vendor/device id of the adapter.
*
Use BUS_PROBE_DEFAULT for pci probe return value
2005-03-05 18:30:12 +00:00
* return BUS_PROBE_DEFAULT on success, positive on failure
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
em_probe(device_t dev)
{
- 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 adapter_name[60];
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 pci_vendor_id = 0;
u16 pci_device_id = 0;
u16 pci_subvendor_id = 0;
u16 pci_subdevice_id = 0;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
em_vendor_info_t *ent;
INIT_DEBUGOUT("em_probe: begin");
pci_vendor_id = pci_get_vendor(dev);
if (pci_vendor_id != EM_VENDOR_ID)
- 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 (ENXIO);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
pci_device_id = pci_get_device(dev);
pci_subvendor_id = pci_get_subvendor(dev);
pci_subdevice_id = pci_get_subdevice(dev);
ent = em_vendor_info_array;
while (ent->vendor_id != 0) {
if ((pci_vendor_id == ent->vendor_id) &&
(pci_device_id == ent->device_id) &&
((pci_subvendor_id == ent->subvendor_id) ||
- 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
(ent->subvendor_id == PCI_ANY_ID)) &&
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
((pci_subdevice_id == ent->subdevice_id) ||
- 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
(ent->subdevice_id == PCI_ANY_ID))) {
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
sprintf(adapter_name, "%s %s",
em_strings[ent->index],
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
em_driver_version);
device_set_desc_copy(dev, adapter_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
return (BUS_PROBE_DEFAULT);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
ent++;
}
- 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 (ENXIO);
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
em_attach(device_t dev)
{
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;
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
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
INIT_DEBUGOUT("em_attach: begin");
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 = device_get_softc(dev);
adapter->dev = adapter->osdep.dev = dev;
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_CORE_LOCK_INIT(adapter, device_get_nameunit(dev));
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)),
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
OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 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
em_sysctl_debug_info, "I", "Debug Information");
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
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
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
OID_AUTO, "stats", CTLTYPE_INT|CTLFLAG_RW, adapter, 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
em_sysctl_stats, "I", "Statistics");
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
callout_init_mtx(&adapter->timer, &adapter->core_mtx, 0);
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 */
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_identify_hardware(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
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 */
if (em_allocate_pci_resources(adapter)) {
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
** must happen after the MAC is
** identified
*/
if ((adapter->hw.mac.type == e1000_ich8lan) ||
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->hw.mac.type == e1000_pchlan) ||
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
(adapter->hw.mac.type == e1000_ich9lan) ||
(adapter->hw.mac.type == e1000_ich10lan)) {
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 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->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
}
/* Do Shared Code initialization */
if (e1000_setup_init_funcs(&adapter->hw, TRUE)) {
device_printf(dev, "Setup of Shared code failed\n");
error = ENXIO;
goto err_pci;
}
e1000_get_bus_info(&adapter->hw);
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,
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_REGISTER(&adapter->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,
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_REGISTER(&adapter->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,
E1000_REGISTER(&adapter->hw, E1000_RADV),
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,
E1000_REGISTER(&adapter->hw, E1000_TADV),
em_tx_abs_int_delay_dflt);
- 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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/* Sysctls for limiting the amount of work done in the taskqueue */
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_add_rx_process_limit(adapter, "rx_processing_limit",
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
"max number of rx packets to process", &adapter->rx_process_limit,
em_rx_process_limit);
- 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
/*
* Validate number of transmit and receive descriptors. It
* must not exceed hardware maximum, and must be multiple
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
* of E1000_DBA_ALIGN.
- 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
*/
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 (((em_txd * sizeof(struct e1000_tx_desc)) % EM_DBA_ALIGN) != 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
(em_txd > EM_MAX_TXD) || (em_txd < EM_MIN_TXD)) {
- 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, "Using %d TX descriptors instead of %d!\n",
EM_DEFAULT_TXD, em_txd);
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->num_tx_desc = EM_DEFAULT_TXD;
- 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
} 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
adapter->num_tx_desc = em_txd;
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
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 (((em_rxd * sizeof(struct e1000_rx_desc)) % EM_DBA_ALIGN) != 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
(em_rxd > EM_MAX_RXD) || (em_rxd < EM_MIN_RXD)) {
- 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, "Using %d RX descriptors instead of %d!\n",
EM_DEFAULT_RXD, em_rxd);
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->num_rx_desc = EM_DEFAULT_RXD;
- 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
} 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
adapter->num_rx_desc = em_rxd;
- 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
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;
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.phy.autoneg_wait_to_complete = FALSE;
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.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
/* Copper options */
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
adapter->hw.phy.mdix = AUTO_ALL_MODES;
adapter->hw.phy.disable_polarity_correction = FALSE;
adapter->hw.phy.ms_type = EM_MASTER_SLAVE;
}
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
*/
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->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
adapter->min_frame_size = ETH_ZLEN + 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
*/
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.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
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
/*
** Get queue/ring memory
*/
if (em_allocate_queues(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
error = ENOMEM;
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_pci;
- 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
}
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 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.
*/
e1000_reset_hw(&adapter->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 */
if (e1000_validate_nvm_checksum(&adapter->hw) < 0) {
/*
** 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.
*/
if (e1000_validate_nvm_checksum(&adapter->hw) < 0) {
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 */
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 (e1000_read_mac_addr(&adapter->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"
" 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
}
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 (!em_is_valid_ether_addr(adapter->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
}
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
/*
** Do interrupt configuration
*/
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 (adapter->msix > 1) /* Do MSIX */
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
error = em_allocate_msix(adapter);
else /* MSI or Legacy */
error = em_allocate_legacy(adapter);
if (error)
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;
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
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
*/
em_get_wakeup(dev);
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 */
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_setup_interface(dev, adapter);
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
em_reset(adapter);
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);
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.get_link_status = 1;
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_link_status(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
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
/* Indicate SOL/IDER usage */
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 (e1000_check_reset_block(&adapter->hw))
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
device_printf(dev,
"PHY reset is blocked due to SOL/IDER session.\n");
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
/* Register for VLAN events */
adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
em_register_vlan, adapter, EVENTHANDLER_PRI_FIRST);
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
adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
em_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST);
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
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);
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
/* Tell the stack that the interface is not active */
adapter->ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
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
adapter->led_dev = led_create(em_led_func, adapter,
device_get_nameunit(dev));
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_attach: end");
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 (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
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
em_free_transmit_structures(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_free_receive_structures(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
em_release_hw_control(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
err_pci:
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_free_pci_resources(adapter);
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_CORE_LOCK_DESTROY(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);
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
em_detach(device_t dev)
{
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 = device_get_softc(dev);
struct ifnet *ifp = adapter->ifp;
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");
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
/* Make sure VLANS are not using driver */
if (adapter->ifp->if_vlantrunk != NULL) {
device_printf(dev,"Vlan in use, detach first\n");
return (EBUSY);
}
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING)
ether_poll_deregister(ifp);
#endif
Fix a mismerge in r206001. PR: 146614 Approved by: jfv (implicit) MFC afer: 3 days
2010-05-15 19:46:16 +00:00
if (adapter->led_dev != NULL)
led_destroy(adapter->led_dev);
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_CORE_LOCK(adapter);
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->in_detach = 1;
em_stop(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_CORE_UNLOCK(adapter);
EM_CORE_LOCK_DESTROY(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
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);
A few small but significant fixes: - Coverity Prevent(tm) CID 1906 a bogus use of bzero where unneeded. - ICH8 systems autoneg to 100 rather than 1000, this can also be seen in 82573, the logic was backwards. - On new 82575 quadports half duplex tx speed is slow... this was due to overwriting TCTL reg rather than adding bits.
2007-05-30 23:32:21 +00:00
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
/* Unregister VLAN events */
if (adapter->vlan_attach != NULL)
EVENTHANDLER_DEREGISTER(vlan_config, adapter->vlan_attach);
if (adapter->vlan_detach != NULL)
EVENTHANDLER_DEREGISTER(vlan_unconfig, adapter->vlan_detach);
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
ether_ifdetach(adapter->ifp);
- Use callout_init_mtx() to close various callout-related races. - Drain the two timers in detach. - Check IFF_DRV_RUNNING in the link task and bail w/o doing anything if it is clear. Reviewed by: jfv, scottl
2006-10-31 17:05:02 +00:00
callout_drain(&adapter->timer);
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_free_pci_resources(adapter);
Only reset the phy when it is absolutely required. This should fix the issues with long *init* times when you do ifconfig em0 alias. MFC after: 3 days
2004-02-10 21:31:09 +00:00
bus_generic_detach(dev);
Make sure that we call if_free(ifp) after bus_teardown_intr. Since we could get an interrupt after we free the ifp, and the interrupt handler depended on the ifp being still alive, this could, in theory, cause a crash. Eliminate this possibility by moving the if_free to after the bus_teardown_intr() call.
2005-09-19 03:10:21 +00:00
if_free(ifp);
Only reset the phy when it is absolutely required. This should fix the issues with long *init* times when you do ifconfig em0 alias. MFC after: 3 days
2004-02-10 21:31:09 +00:00
Allocate receive and transmit data structures during attach() and free them during detach() similar to other NIC drivers rather than allocating them during init() and freeing them during stop(): - Move creation of tx bus_dma tag amd maps and tx_buffer_area from em_setup_transmit_structures() to em_allocate_transmit_structures(). - Call em_allocate_xxx_structures() in em_attach(). - Only call em_free_xxx_structures() in em_detach(). - Change em_setup_xxx_structures() to free any existing tx or rx buffers and in the case of rx repopulate the ring with newer buffers. Reviewed by: jfv
2006-10-31 17:21:14 +00:00
em_free_transmit_structures(adapter);
em_free_receive_structures(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);
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
em_shutdown(device_t 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
return em_suspend(dev);
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
em_suspend(device_t dev)
{
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 = device_get_softc(dev);
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
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_CORE_LOCK(adapter);
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
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_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);
em_enable_wakeup(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
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_CORE_UNLOCK(adapter);
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
return bus_generic_suspend(dev);
}
static int
em_resume(device_t dev)
{
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 = device_get_softc(dev);
struct ifnet *ifp = adapter->ifp;
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
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_CORE_LOCK(adapter);
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_init_locked(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
em_init_manageability(adapter);
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_CORE_UNLOCK(adapter);
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_start(ifp);
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
return bus_generic_resume(dev);
}
- 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
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
/*********************************************************************
* Transmit entry point
*
* em_start is called by the stack to initiate a transmit.
* The driver will remain in this routine as long as there are
* packets to transmit and transmit resources are available.
* In case resources are not available stack is notified and
* the packet is requeued.
**********************************************************************/
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
#ifdef EM_MULTIQUEUE
- 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 int
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_mq_start_locked(struct ifnet *ifp, struct tx_ring *txr, struct mbuf *m)
- 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
{
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 adapter *adapter = txr->adapter;
struct mbuf *next;
int err = 0, enq = 0;
if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
IFF_DRV_RUNNING || adapter->link_active == 0) {
if (m != NULL)
err = drbr_enqueue(ifp, txr->br, m);
return (err);
}
- 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
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
/* Call cleanup if number of TX descriptors low */
if (txr->tx_avail <= EM_TX_CLEANUP_THRESHOLD)
em_txeof(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
enq = 0;
Three changes: - add CRC stripping to the RX side, this was handled by some obscure code in rxeof previously, its easier to simply have the hardware strip it now. - Add back an ALTQ change that slipped between the cracks - Add an update to the watchdog_time in the xmit code, not doing this in ixgbe caused problems, think its needed here as well.
2010-04-08 19:13:42 +00:00
if (m == NULL) {
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
next = drbr_dequeue(ifp, txr->br);
Three changes: - add CRC stripping to the RX side, this was handled by some obscure code in rxeof previously, its easier to simply have the hardware strip it now. - Add back an ALTQ change that slipped between the cracks - Add an update to the watchdog_time in the xmit code, not doing this in ixgbe caused problems, think its needed here as well.
2010-04-08 19:13:42 +00:00
} else if (drbr_needs_enqueue(ifp, txr->br)) {
if ((err = drbr_enqueue(ifp, txr->br, m)) != 0)
return (err);
next = drbr_dequeue(ifp, txr->br);
} 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
next = m;
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
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
/* Process the queue */
while (next != NULL) {
if ((err = em_xmit(txr, &next)) != 0) {
if (next != NULL)
err = drbr_enqueue(ifp, txr->br, next);
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
break;
Fix for pr 138516 An mbuf is not requeued when a xmit fails. MFC: 3 days
2009-09-10 18:53:34 +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
enq++;
Fix for pr 138516 An mbuf is not requeued when a xmit fails. MFC: 3 days
2009-09-10 18:53:34 +00:00
drbr_stats_update(ifp, next->m_pkthdr.len, next->m_flags);
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
ETHER_BPF_MTAP(ifp, next);
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
break;
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
if (txr->tx_avail < EM_MAX_SCATTER) {
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
break;
}
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
next = drbr_dequeue(ifp, txr->br);
}
- 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
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 (enq > 0) {
/* Set the watchdog */
txr->watchdog_check = TRUE;
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
txr->watchdog_time = ticks;
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 (err);
- 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
}
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
/*
** Multiqueue capable stack interface, this is not
** yet truely multiqueue, but that is coming...
*/
- 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 int
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
em_mq_start(struct ifnet *ifp, struct mbuf *m)
- 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
{
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 adapter *adapter = ifp->if_softc;
struct tx_ring *txr;
int i, error = 0;
- 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
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
/* Which queue to use */
if ((m->m_flags & M_FLOWID) != 0)
i = m->m_pkthdr.flowid % adapter->num_queues;
else
i = curcpu % adapter->num_queues;
txr = &adapter->tx_rings[i];
if (EM_TX_TRYLOCK(txr)) {
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
error = em_mq_start_locked(ifp, txr, m);
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_UNLOCK(txr);
- 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
} 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
error = drbr_enqueue(ifp, txr->br, m);
- 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
return (error);
}
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
/*
** Flush all ring buffers
*/
- 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
em_qflush(struct ifnet *ifp)
{
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 adapter *adapter = ifp->if_softc;
struct tx_ring *txr = adapter->tx_rings;
struct mbuf *m;
- 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
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
for (int i = 0; i < adapter->num_queues; i++, txr++) {
EM_TX_LOCK(txr);
while ((m = buf_ring_dequeue_sc(txr->br)) != NULL)
m_freem(m);
EM_TX_UNLOCK(txr);
}
call base if_qflush routine to flush if_snd
2009-04-16 23:08:02 +00:00
if_qflush(ifp);
- 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
}
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
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
#endif /* EM_MULTIQUEUE */
- 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
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_start_locked(struct ifnet *ifp, struct tx_ring *txr)
- 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
{
struct adapter *adapter = ifp->if_softc;
struct mbuf *m_head;
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_LOCK_ASSERT(txr);
- 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
if ((ifp->if_drv_flags & (IFF_DRV_RUNNING|IFF_DRV_OACTIVE)) !=
IFF_DRV_RUNNING)
return;
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
- 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
if (!adapter->link_active)
return;
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
/* Call cleanup if number of TX descriptors low */
if (txr->tx_avail <= EM_TX_CLEANUP_THRESHOLD)
em_txeof(txr);
- 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
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) {
if (txr->tx_avail < EM_MAX_SCATTER) {
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
break;
}
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
IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
- 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
if (m_head == NULL)
break;
/*
* Encapsulation can modify our pointer, and or make it
* NULL on failure. In that event, we can't requeue.
*/
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_xmit(txr, &m_head)) {
- 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
if (m_head == NULL)
break;
Thanks for Michael Tuexen for tracking down a path where the watchdog timer was not being rearmed in txeof, and also a missing case in the new code. MFC after: 2 weeks
2009-04-27 17:22:14 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
- 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
break;
}
/* Send a copy of the frame to the BPF listener */
ETHER_BPF_MTAP(ifp, m_head);
/* Set timeout in case hardware has problems transmitting. */
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
txr->watchdog_time = ticks;
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
txr->watchdog_check = TRUE;
- 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
}
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
return;
- 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
}
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
static void
em_start(struct ifnet *ifp)
{
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 adapter *adapter = ifp->if_softc;
struct tx_ring *txr = adapter->tx_rings;
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +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 (ifp->if_drv_flags & IFF_DRV_RUNNING) {
EM_TX_LOCK(txr);
em_start_locked(ifp, txr);
EM_TX_UNLOCK(txr);
}
return;
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +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
/*********************************************************************
* Ioctl entry point
*
* em_ioctl is called when the user wants to configure the
* interface.
*
* return 0 on success, positive on failure
**********************************************************************/
static int
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
em_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
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
{
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 = ifp->if_softc;
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
struct ifreq *ifr = (struct ifreq *)data;
Hide AF_INET specific ioctl handling under #ifdef INET. MFC after: 2 months
2008-11-06 11:00:57 +00:00
#ifdef INET
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +00:00
struct ifaddr *ifa = (struct ifaddr *)data;
Hide AF_INET specific ioctl handling under #ifdef INET. MFC after: 2 months
2008-11-06 11:00:57 +00:00
#endif
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
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
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->in_detach)
- 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);
- Code cleanup - In the receive routine handle the case where last descriptor could have less than 4 bytes of data. - Handle race between detach/ioctl routine. MFC after: 3 days
2003-11-14 18:02:25 +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
switch (command) {
case SIOCSIFADDR:
Hide AF_INET specific ioctl handling under #ifdef INET. MFC after: 2 months
2008-11-06 11:00:57 +00:00
#ifdef INET
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +00:00
if (ifa->ifa_addr->sa_family == AF_INET) {
/*
* XXX
* Since resetting hardware takes a very long time
* and results in link renegotiation we only
* initialize the hardware only when it is absolutely
* required.
*/
ifp->if_flags |= IFF_UP;
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
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_CORE_LOCK(adapter);
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_init_locked(adapter);
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_CORE_UNLOCK(adapter);
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +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
arp_ifinit(ifp, ifa);
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +00:00
} else
Hide AF_INET specific ioctl handling under #ifdef INET. MFC after: 2 months
2008-11-06 11:00:57 +00:00
#endif
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +00:00
error = ether_ioctl(ifp, command, data);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
break;
case SIOCSIFMTU:
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
{
int max_frame_size;
IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
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_CORE_LOCK(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
switch (adapter->hw.mac.type) {
case e1000_82571:
case e1000_82572:
case e1000_ich9lan:
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
case e1000_ich10lan:
case e1000_82574:
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
case e1000_80003es2lan: /* 9K Jumbo Frame size */
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
max_frame_size = 9234;
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
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_pchlan:
max_frame_size = 4096;
break;
Two minor fixes, keep old 82542 from using jumbo frames, and add missing htole64 in encap code. Reviewed by:Pdeuskar Approved by:Pdeuskar
2007-05-23 20:41:20 +00:00
/* Adapters that do not support jumbo frames */
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
case e1000_82583:
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
case e1000_ich8lan:
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
max_frame_size = ETHER_MAX_LEN;
break;
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
default:
max_frame_size = MAX_JUMBO_FRAME_SIZE;
}
if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN -
ETHER_CRC_LEN) {
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_CORE_UNLOCK(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
error = EINVAL;
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
break;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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
ifp->if_mtu = ifr->ifr_mtu;
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->max_frame_size =
ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
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_init_locked(adapter);
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_CORE_UNLOCK(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
break;
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
}
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
case SIOCSIFFLAGS:
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
IOCTL_DEBUGOUT("ioctl rcv'd:\
SIOCSIFFLAGS (Set Interface Flags)");
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_CORE_LOCK(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
if (ifp->if_flags & IFF_UP) {
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING)) {
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 ((ifp->if_flags ^ adapter->if_flags) &
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
(IFF_PROMISC | IFF_ALLMULTI)) {
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_disable_promisc(adapter);
em_set_promisc(adapter);
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +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
em_init_locked(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
} 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
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
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_stop(adapter);
adapter->if_flags = ifp->if_flags;
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_CORE_UNLOCK(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
break;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
case SIOCADDMULTI:
case SIOCDELMULTI:
IOCTL_DEBUGOUT("ioctl rcv'd: SIOC(ADD|DEL)MULTI");
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
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_CORE_LOCK(adapter);
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_disable_intr(adapter);
em_set_multi(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
#ifdef DEVICE_POLLING
- 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 (!(ifp->if_capenable & IFCAP_POLLING))
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
#endif
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_enable_intr(adapter);
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_CORE_UNLOCK(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
break;
case SIOCSIFMEDIA:
Merge new vendor release - 6.2.9. Details: o if_em.c changes: - Added several new PCI ids. - Check em_check_phy_reset_block() before doing SIOCSIFMEDIA ioctl. - Don't touch TARC registers, they are now handled in shared code in if_em_hw.c. - Move RDH and RDT setting to the end of em_initialize_receive_unit(). - Declare em_read_pcie_cap_reg(), now empty. o if_em_hw.c dropped in from vendor, then restored rev. 1.15. o if_em_hw.h dropped in from vendor, then modified: - Added RX overrun interrupt flag to interrupt enable mask. - Remove declarations of em_io_read(), em_io_write(). Approved by: jfv
2006-10-31 15:00:14 +00:00
/* Check SOL/IDER usage */
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_CORE_LOCK(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
if (e1000_check_reset_block(&adapter->hw)) {
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_CORE_UNLOCK(adapter);
Merge new vendor release - 6.2.9. Details: o if_em.c changes: - Added several new PCI ids. - Check em_check_phy_reset_block() before doing SIOCSIFMEDIA ioctl. - Don't touch TARC registers, they are now handled in shared code in if_em_hw.c. - Move RDH and RDT setting to the end of em_initialize_receive_unit(). - Declare em_read_pcie_cap_reg(), now empty. o if_em_hw.c dropped in from vendor, then restored rev. 1.15. o if_em_hw.h dropped in from vendor, then modified: - Added RX overrun interrupt flag to interrupt enable mask. - Remove declarations of em_io_read(), em_io_write(). Approved by: jfv
2006-10-31 15:00:14 +00:00
device_printf(adapter->dev, "Media change 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
" blocked due to SOL/IDER session.\n");
Merge new vendor release - 6.2.9. Details: o if_em.c changes: - Added several new PCI ids. - Check em_check_phy_reset_block() before doing SIOCSIFMEDIA ioctl. - Don't touch TARC registers, they are now handled in shared code in if_em_hw.c. - Move RDH and RDT setting to the end of em_initialize_receive_unit(). - Declare em_read_pcie_cap_reg(), now empty. o if_em_hw.c dropped in from vendor, then restored rev. 1.15. o if_em_hw.h dropped in from vendor, then modified: - Added RX overrun interrupt flag to interrupt enable mask. - Remove declarations of em_io_read(), em_io_write(). Approved by: jfv
2006-10-31 15:00:14 +00:00
break;
}
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_CORE_UNLOCK(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
case SIOCGIFMEDIA:
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
IOCTL_DEBUGOUT("ioctl rcv'd: \
SIOCxIFMEDIA (Get/Set Interface Media)");
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
error = ifmedia_ioctl(ifp, ifr, &adapter->media, command);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
break;
case SIOCSIFCAP:
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
{
int mask, reinit;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFCAP (Set Capabilities)");
Implement support for controlling VLAN_HWTAGGING through ioctl(SIOCSIFCAP). This includes not only toggling the flag in if_capenable, but also really reconfiguring the hardware. Approved by: tackerman (as the em(4) maintainer)
2004-06-04 16:57:18 +00:00
reinit = 0;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
mask = ifr->ifr_reqcap ^ ifp->if_capenable;
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
#ifdef DEVICE_POLLING
if (mask & IFCAP_POLLING) {
if (ifr->ifr_reqcap & IFCAP_POLLING) {
error = ether_poll_register(em_poll, ifp);
if (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
return (error);
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_CORE_LOCK(adapter);
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_disable_intr(adapter);
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
ifp->if_capenable |= IFCAP_POLLING;
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_CORE_UNLOCK(adapter);
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
} else {
error = ether_poll_deregister(ifp);
/* Enable interrupt even in error case */
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_CORE_LOCK(adapter);
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_enable_intr(adapter);
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
ifp->if_capenable &= ~IFCAP_POLLING;
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_CORE_UNLOCK(adapter);
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
}
}
#endif
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
if (mask & IFCAP_HWCSUM) {
Implemented per-interface polling(4) control.
2004-04-11 21:01:12 +00:00
ifp->if_capenable ^= IFCAP_HWCSUM;
Implement support for controlling VLAN_HWTAGGING through ioctl(SIOCSIFCAP). This includes not only toggling the flag in if_capenable, but also really reconfiguring the hardware. Approved by: tackerman (as the em(4) maintainer)
2004-06-04 16:57:18 +00:00
reinit = 1;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (mask & IFCAP_TSO4) {
ifp->if_capenable ^= IFCAP_TSO4;
reinit = 1;
}
Implement support for controlling VLAN_HWTAGGING through ioctl(SIOCSIFCAP). This includes not only toggling the flag in if_capenable, but also really reconfiguring the hardware. Approved by: tackerman (as the em(4) maintainer)
2004-06-04 16:57:18 +00:00
if (mask & IFCAP_VLAN_HWTAGGING) {
ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
reinit = 1;
}
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
if (mask & IFCAP_VLAN_HWFILTER) {
ifp->if_capenable ^= IFCAP_VLAN_HWFILTER;
reinit = 1;
}
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 ((mask & IFCAP_WOL) &&
(ifp->if_capabilities & IFCAP_WOL) != 0) {
if (mask & IFCAP_WOL_MCAST)
ifp->if_capenable ^= IFCAP_WOL_MCAST;
if (mask & IFCAP_WOL_MAGIC)
ifp->if_capenable ^= IFCAP_WOL_MAGIC;
}
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (reinit && (ifp->if_drv_flags & IFF_DRV_RUNNING))
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_init(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
VLAN_CAPABILITIES(ifp);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
break;
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
}
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
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
default:
Since resetting hardware takes a very long time and results in link renegotiation, we only initialize the hardware only when it is absolutely required. Process SIOCGIFADDR ioctl in em(4) when we know an IPv4 address is added. Handling SIOCGIFADDR in a driver is layering violation but it seems that there is no easy way without rewritting hardware initialization code to reduce settle time after reset. This should fix a long standing bug which didn't send ARP packet when interface address is changed or an alias address is added. Another effect of this fix is it doesn't need additional delays anymore when adding an alias address to the interface. While I'm here add a new if_flags into softc which remembers current prgroammed interface flags and make use of it when we have to program promiscuous mode. Tested by: Atanas <atanas AT asd DOT aplus DOT net> Analyzed by: rwatson Discussed with: -stable
2006-07-20 04:18:45 +00:00
error = ether_ioctl(ifp, command, data);
break;
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
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
}
/*********************************************************************
* 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
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
* by the driver as a hw/sw initialization routine to get to a
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
* consistent state.
*
* return 0 on success, positive on failure
**********************************************************************/
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_init_locked(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
{
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 ifnet *ifp = adapter->ifp;
device_t dev = adapter->dev;
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 pba;
Implement support for controlling VLAN_HWTAGGING through ioctl(SIOCSIFCAP). This includes not only toggling the flag in if_capenable, but also really reconfiguring the hardware. Approved by: tackerman (as the em(4) maintainer)
2004-06-04 16:57:18 +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_init: begin");
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_CORE_LOCK_ASSERT(adapter);
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
em_disable_intr(adapter);
callout_stop(&adapter->timer);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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
/*
* Packet Buffer Allocation (PBA)
Added support for Intel PRO/1000 GT Desktop Adapter(Device ID 8086 107C) Removed support for Intel 82541ER Added fix for 82547 which corrects an issue with Jumbo frames larger than 10k. Added fix for vlan tagged frames not being properly bridged. Corrected TBI workaround. Corrected incorrect LED operation issues Submitted by: tackerman (Tony Ackerman) MFC after: 2 weeks
2004-09-01 23:22:41 +00:00
* Writing PBA sets the receive portion of the buffer
* the remainder is used for the transmit buffer.
*/
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) {
/* Total Packet Buffer on these is 48K */
case e1000_82571:
case e1000_82572:
case e1000_80003es2lan:
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
pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
break;
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
case e1000_82573: /* 82573: Total Packet Buffer is 32K */
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
pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
break;
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
case e1000_82574:
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
case e1000_82583:
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
pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
break;
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
case e1000_ich9lan:
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
case e1000_ich10lan:
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_pchlan:
pba = E1000_PBA_10K;
break;
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
case e1000_ich8lan:
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
pba = E1000_PBA_8K;
break;
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
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->max_frame_size > 8192)
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
pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
else
pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
Added support for Intel PRO/1000 GT Desktop Adapter(Device ID 8086 107C) Removed support for Intel 82541ER Added fix for 82547 which corrects an issue with Jumbo frames larger than 10k. Added fix for vlan tagged frames not being properly bridged. Corrected TBI workaround. Corrected incorrect LED operation issues Submitted by: tackerman (Tony Ackerman) MFC after: 2 weeks
2004-09-01 23:22:41 +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
Added support for Intel PRO/1000 GT Desktop Adapter(Device ID 8086 107C) Removed support for Intel 82541ER Added fix for 82547 which corrects an issue with Jumbo frames larger than 10k. Added fix for vlan tagged frames not being properly bridged. Corrected TBI workaround. Corrected incorrect LED operation issues Submitted by: tackerman (Tony Ackerman) MFC after: 2 weeks
2004-09-01 23:22:41 +00:00
INIT_DEBUGOUT1("em_init: pba=%dK",pba);
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_PBA, pba);
Added support for Intel PRO/1000 GT Desktop Adapter(Device ID 8086 107C) Removed support for Intel 82541ER Added fix for 82547 which corrects an issue with Jumbo frames larger than 10k. Added fix for vlan tagged frames not being properly bridged. Corrected TBI workaround. Corrected incorrect LED operation issues Submitted by: tackerman (Tony Ackerman) MFC after: 2 weeks
2004-09-01 23:22:41 +00:00
- Code cleanup - In the receive routine handle the case where last descriptor could have less than 4 bytes of data. - Handle race between detach/ioctl routine. MFC after: 3 days
2003-11-14 18:02:25 +00:00
/* Get the latest mac address, User can use a LAA */
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
bcopy(IF_LLADDR(adapter->ifp), adapter->hw.mac.addr,
ETHER_ADDR_LEN);
Back out 1.112,1.113. I don't have enough resources to fix breakages introduced by this change.
2006-02-22 14:11:16 +00:00
A couple bug fixes that I've had internally at Intel. First is a long time workaround for problems with 82571 adapters and LAAs, one port getting reset can cause the other to have its RAR[0] also reset, thus overwriting an LAA. This fix works around it by also keeping the address in the last array member. The other bug is specific to the new 575 adapter, its transmit code logic in handling hwassists was too crude, it broken when doing bridges. I am much happier with the new logic,we may want to change the legacy path at some point to something similar. Reviewed by: pdeuskar Approved by: pdeuskar
2007-05-09 00:41:45 +00:00
/* Put the address into the Receive Address Array */
e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 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
* 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.
A couple bug fixes that I've had internally at Intel. First is a long time workaround for problems with 82571 adapters and LAAs, one port getting reset can cause the other to have its RAR[0] also reset, thus overwriting an LAA. This fix works around it by also keeping the address in the last array member. The other bug is specific to the new 575 adapter, its transmit code logic in handling hwassists was too crude, it broken when doing bridges. I am much happier with the new logic,we may want to change the legacy path at some point to something similar. Reviewed by: pdeuskar Approved by: pdeuskar
2007-05-09 00:41:45 +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.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);
}
A couple bug fixes that I've had internally at Intel. First is a long time workaround for problems with 82571 adapters and LAAs, one port getting reset can cause the other to have its RAR[0] also reset, thus overwriting an LAA. This fix works around it by also keeping the address in the last array member. The other bug is specific to the new 575 adapter, its transmit code logic in handling hwassists was too crude, it broken when doing bridges. I am much happier with the new logic,we may want to change the legacy path at some point to something similar. Reviewed by: pdeuskar Approved by: pdeuskar
2007-05-09 00:41:45 +00:00
Back out 1.112,1.113. I don't have enough resources to fix breakages introduced by this change.
2006-02-22 14:11:16 +00:00
/* Initialize the hardware */
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_reset(adapter);
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_link_status(adapter);
- Code cleanup - In the receive routine handle the case where last descriptor could have less than 4 bytes of data. - Handle race between detach/ioctl routine. MFC after: 3 days
2003-11-14 18:02:25 +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
/* Setup VLAN support, basic and offload if available */
E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
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
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
/* Use real VLAN Filter support? */
if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) {
if (ifp->if_capenable & IFCAP_VLAN_HWFILTER)
/* 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);
}
}
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
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
/* Set hardware offload abilities */
Move the initialization of the hardware capabilities in em_init_locked() before em_setup_transmit_structures() as it needs this information to properly set up TSO parameters. Reviewed by: jfv
2006-09-21 19:14:51 +00:00
ifp->if_hwassist = 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
if (ifp->if_capenable & IFCAP_TXCSUM)
ifp->if_hwassist |= (CSUM_TCP | CSUM_UDP);
if (ifp->if_capenable & IFCAP_TSO4)
ifp->if_hwassist |= CSUM_TSO;
Move the initialization of the hardware capabilities in em_init_locked() before em_setup_transmit_structures() as it needs this information to properly set up TSO parameters. Reviewed by: jfv
2006-09-21 19:14:51 +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
/* Configure for OS presence */
em_init_manageability(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Prepare transmit descriptors and buffers */
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_setup_transmit_structures(adapter);
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_initialize_transmit_unit(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Setup Multicast table */
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_set_multi(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Prepare receive descriptors and buffers */
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 (em_setup_receive_structures(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
device_printf(dev, "Could not setup receive structures\n");
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_stop(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
return;
}
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_initialize_receive_unit(adapter);
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
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 */
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_set_promisc(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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
callout_reset(&adapter->timer, hz, em_local_timer, 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
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) {
int tmp;
tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
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);
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 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
Add polling support to the "em" driver. Reviewed by: pdeuskar (maintainer) Approved by: re
2002-11-25 19:56:16 +00:00
#ifdef DEVICE_POLLING
- 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
/*
* Only enable interrupts if we are not polling, make sure
* they are off otherwise.
*/
if (ifp->if_capenable & IFCAP_POLLING)
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_disable_intr(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
else
Add polling support to the "em" driver. Reviewed by: pdeuskar (maintainer) Approved by: re
2002-11-25 19:56:16 +00:00
#endif /* DEVICE_POLLING */
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_enable_intr(adapter);
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);
Only reset the phy when it is absolutely required. This should fix the issues with long *init* times when you do ifconfig em0 alias. MFC after: 3 days
2004-02-10 21:31:09 +00:00
/* Don't reset the phy next time init gets called */
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.phy.reset_disable = TRUE;
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
}
static void
em_init(void *arg)
{
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 = arg;
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +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
EM_CORE_LOCK(adapter);
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_init_locked(adapter);
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_CORE_UNLOCK(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 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
Add polling support to the "em" driver. Reviewed by: pdeuskar (maintainer) Approved by: re
2002-11-25 19:56:16 +00:00
#ifdef DEVICE_POLLING
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
* Legacy polling routine: note this only works with single queue
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
*
*********************************************************************/
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int
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
em_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
Add polling support to the "em" driver. Reviewed by: pdeuskar (maintainer) Approved by: re
2002-11-25 19:56:16 +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
struct adapter *adapter = ifp->if_softc;
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 tx_ring *txr = adapter->tx_rings;
struct rx_ring *rxr = adapter->rx_rings;
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
u32 reg_icr;
int rx_done;
- 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
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_CORE_LOCK(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
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 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
EM_CORE_UNLOCK(adapter);
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
return (0);
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
}
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +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
if (cmd == POLL_AND_CHECK_STATUS) {
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);
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 (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
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
callout_stop(&adapter->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
adapter->hw.mac.get_link_status = 1;
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_link_status(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
callout_reset(&adapter->timer, hz,
em_local_timer, adapter);
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
}
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_CORE_UNLOCK(adapter);
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
em_rxeof(rxr, count, &rx_done);
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
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_LOCK(txr);
em_txeof(txr);
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
#ifdef EM_MULTIQUEUE
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 (!drbr_empty(ifp, txr->br))
em_mq_start_locked(ifp, txr, NULL);
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
#else
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 (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
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_start_locked(ifp, txr);
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
#endif
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_UNLOCK(txr);
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
return (rx_done);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +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
#endif /* DEVICE_POLLING */
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 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
/*********************************************************************
*
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
* Fast Legacy/MSI Combined Interrupt Service routine
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
*
*********************************************************************/
o break newbus api: add a new argument of type driver_filter_t to bus_setup_intr() o add an int return code to all fast handlers o retire INTR_FAST/IH_FAST For more info: http://docs.freebsd.org/cgi/getmsg.cgi?fetch=465712+0+current/freebsd-current Reviewed by: many Approved by: re@
2007-02-23 12:19:07 +00:00
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_irq_fast(void *arg)
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
{
struct adapter *adapter = arg;
struct ifnet *ifp;
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_icr;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
ifp = adapter->ifp;
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
/* Hot eject? */
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
/* Definitely not our interrupt. */
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
em_disable_intr(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
taskqueue_enqueue(adapter->tq, &adapter->que_task);
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;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
taskqueue_enqueue(taskqueue_fast, &adapter->link_task);
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++;
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_HANDLED;
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
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
/* Combined RX/TX handler, used by Legacy and MSI */
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
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_handle_que(void *context, int pending)
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
struct adapter *adapter = context;
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 ifnet *ifp = adapter->ifp;
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 tx_ring *txr = adapter->tx_rings;
struct rx_ring *rxr = adapter->rx_rings;
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
bool more;
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
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
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
more = em_rxeof(rxr, adapter->rx_process_limit, NULL);
The lock move in rxeof necessitated a couple more places to do the locking, fixes a panic.
2010-04-10 19:25:55 +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
EM_TX_LOCK(txr);
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
if (em_txeof(txr))
more = TRUE;
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
#ifdef EM_MULTIQUEUE
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 (!drbr_empty(ifp, txr->br))
em_mq_start_locked(ifp, txr, NULL);
#else
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
em_start_locked(ifp, txr);
#endif
EM_TX_UNLOCK(txr);
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
if (more) {
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
taskqueue_enqueue(adapter->tq, &adapter->que_task);
A few more changes from yongari: - code flow in handler could let interrupt be reenabled when not wanted. - change where the RX lock is taken to improve performance. - adapter->msix is true for MSI systems also, it needs to explicitly test for 82574, good one :)
2010-04-09 23:15:37 +00:00
return;
}
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
}
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_enable_intr(adapter);
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
return;
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
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
* MSIX Interrupt Service Routines
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
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_msix_tx(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
{
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 tx_ring *txr = arg;
struct adapter *adapter = txr->adapter;
Add a missing fragment in the tx msix handler to invoke another if all work is not done. Sync the igb driver with changes suggested by yongari and made in em, these made sense to be in both drivers.
2010-04-14 20:55:33 +00:00
bool more;
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
++txr->tx_irq;
EM_TX_LOCK(txr);
Add a missing fragment in the tx msix handler to invoke another if all work is not done. Sync the igb driver with changes suggested by yongari and made in em, these made sense to be in both drivers.
2010-04-14 20:55:33 +00:00
more = em_txeof(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
EM_TX_UNLOCK(txr);
Add a missing fragment in the tx msix handler to invoke another if all work is not done. Sync the igb driver with changes suggested by yongari and made in em, these made sense to be in both drivers.
2010-04-14 20:55:33 +00:00
if (more)
taskqueue_enqueue(txr->tq, &txr->tx_task);
else
/* Reenable this interrupt */
E1000_WRITE_REG(&adapter->hw, E1000_IMS, txr->ims);
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;
}
/*********************************************************************
*
* MSIX RX Interrupt Service routine
*
**********************************************************************/
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
static void
em_msix_rx(void *arg)
{
struct rx_ring *rxr = arg;
struct adapter *adapter = rxr->adapter;
bool more;
++rxr->rx_irq;
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
more = em_rxeof(rxr, adapter->rx_process_limit, NULL);
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 (more)
taskqueue_enqueue(rxr->tq, &rxr->rx_task);
else
/* Reenable this interrupt */
E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxr->ims);
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
return;
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
*
**********************************************************************/
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
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
{
struct adapter *adapter = arg;
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
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;
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
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
if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
adapter->hw.mac.get_link_status = 1;
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
em_handle_link(adapter, 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
} else
E1000_WRITE_REG(&adapter->hw, E1000_IMS,
EM_MSIX_LINK | 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
return;
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 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
static void
em_handle_rx(void *context, int pending)
{
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 rx_ring *rxr = context;
struct adapter *adapter = rxr->adapter;
bool more;
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
more = em_rxeof(rxr, adapter->rx_process_limit, NULL);
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
if (more)
taskqueue_enqueue(rxr->tq, &rxr->rx_task);
else
/* Reenable this interrupt */
E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxr->ims);
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
}
static void
em_handle_tx(void *context, int pending)
{
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 tx_ring *txr = context;
struct adapter *adapter = txr->adapter;
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
struct ifnet *ifp = adapter->ifp;
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_TX_TRYLOCK(txr))
return;
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
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
em_txeof(txr);
#ifdef EM_MULTIQUEUE
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 (!drbr_empty(ifp, txr->br))
em_mq_start_locked(ifp, txr, NULL);
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
#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
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
em_start_locked(ifp, txr);
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
#endif
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, txr->ims);
EM_TX_UNLOCK(txr);
}
static void
em_handle_link(void *context, int pending)
{
struct adapter *adapter = context;
struct ifnet *ifp = adapter->ifp;
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING))
return;
EM_CORE_LOCK(adapter);
callout_stop(&adapter->timer);
em_update_link_status(adapter);
callout_reset(&adapter->timer, hz, em_local_timer, adapter);
E1000_WRITE_REG(&adapter->hw, E1000_IMS,
EM_MSIX_LINK | E1000_IMS_LSC);
EM_CORE_UNLOCK(adapter);
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
em_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
{
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 = ifp->if_softc;
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;
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_media_status: begin");
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_CORE_LOCK(adapter);
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_link_status(adapter);
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) {
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_CORE_UNLOCK(adapter);
- 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)) {
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;
}
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_CORE_UNLOCK(adapter);
- 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
}
/*********************************************************************
*
* Media Ioctl callback
*
* This routine is called when the user changes speed/duplex using
* media/mediopt option with ifconfig.
*
**********************************************************************/
static int
em_media_change(struct ifnet *ifp)
{
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 = ifp->if_softc;
struct ifmedia *ifm = &adapter->media;
- 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
INIT_DEBUGOUT("em_media_change: begin");
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
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_CORE_LOCK(adapter);
- 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
}
Only reset the phy when it is absolutely required. This should fix the issues with long *init* times when you do ifconfig em0 alias. MFC after: 3 days
2004-02-10 21:31:09 +00:00
/* As the speed/duplex settings my have changed we need to
* reset the PHY.
*/
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.phy.reset_disable = FALSE;
Only reset the phy when it is absolutely required. This should fix the issues with long *init* times when you do ifconfig em0 alias. MFC after: 3 days
2004-02-10 21:31:09 +00:00
Add locking to the ifmedia callouts. Reviewed by: jvogel, yongari
2006-09-01 16:08:36 +00:00
em_init_locked(adapter);
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_CORE_UNLOCK(adapter);
- 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
}
/*********************************************************************
*
* This routine maps the mbufs to tx descriptors.
*
* return 0 on success, positive on failure
**********************************************************************/
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
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
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_xmit(struct tx_ring *txr, struct mbuf **m_headp)
- 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
struct adapter *adapter = txr->adapter;
Convert if_em to the new bus_dmamap_load_sg() interface. The old callback was really just a waste of cycles, so this streamlines it considerably.
2005-01-15 20:52:15 +00:00
bus_dma_segment_t segs[EM_MAX_SCATTER];
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
bus_dmamap_t map;
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
struct em_buffer *tx_buffer, *tx_buffer_mapped;
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
struct e1000_tx_desc *ctxd = NULL;
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
struct mbuf *m_head;
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 txd_upper, txd_lower, txd_used, txd_saved;
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
int nsegs, i, j, first, last = 0;
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +00:00
int error, do_tso, tso_desc = 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
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
m_head = *m_headp;
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +00:00
txd_upper = txd_lower = txd_used = txd_saved = 0;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
do_tso = ((m_head->m_pkthdr.csum_flags & CSUM_TSO) != 0);
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
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +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
* TSO workaround:
* If an mbuf is only header we need
* to pull 4 bytes of data into it.
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
*/
if (do_tso && (m_head->m_len <= M_TSO_LEN)) {
m_head = m_pullup(m_head, M_TSO_LEN + 4);
Fix issues found by Coverity (223392, 223393) due to TSO additions Submitted by: Matthew Jacob
2006-09-11 20:59:01 +00:00
*m_headp = m_head;
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 (m_head == NULL)
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
return (ENOBUFS);
}
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +00:00
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
* Map the packet for DMA
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
*
* Capture the first descriptor index,
* this descriptor will have the index
* of the EOP which is the only one that
* now gets a DONE bit writeback.
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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 = txr->next_avail_desc;
tx_buffer = &txr->tx_buffers[first];
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
tx_buffer_mapped = tx_buffer;
Prepending an mbuf after loading a DMA map results in unexpected result. So, modify mbuf chains before loading a DMA map.
2006-07-27 00:43:34 +00:00
map = tx_buffer->map;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
error = bus_dmamap_load_mbuf_sg(txr->txtag, map,
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
*m_headp, segs, &nsegs, BUS_DMA_NOWAIT);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*
* There are two types of errors we can (try) to handle:
* - EFBIG means the mbuf chain was too long and bus_dma ran
* out of segments. Defragment the mbuf chain and try again.
* - ENOMEM means bus_dma could not obtain enough bounce buffers
* at this point in time. Defer sending and try again later.
* All other errors, in particular EINVAL, are fatal and prevent the
* mbuf chain from ever going through. Drop it and report error.
*/
Make em(4) handle too many fragmented frame with m_defrag(9). Previously em(4) requeued the failed mbuf chains from bus_dmamap_load_mbuf_sg(9) failure to resend it later. However, bus_dmamap_load_mbuf_sg(9) may never complete its request as the fragmented frames can have more than EM_MAX_SCATTER segments. To handle the above EFBIG case, defragment the frame with m_defrag(9) and free the mbuf chain if it can't deframent the chain due to resource shortage. Reviewed by glebius (with improvements)
2006-08-14 02:21:26 +00:00
if (error == EFBIG) {
struct mbuf *m;
m = m_defrag(*m_headp, M_DONTWAIT);
if (m == NULL) {
adapter->mbuf_alloc_failed++;
m_freem(*m_headp);
*m_headp = NULL;
return (ENOBUFS);
}
*m_headp = m;
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
/* Try it again */
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 = bus_dmamap_load_mbuf_sg(txr->txtag, map,
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
*m_headp, segs, &nsegs, BUS_DMA_NOWAIT);
Handle all error cases from bus_dmamap_load_mbuf_sg(). Those are: - EFBIG means the mbuf chain was too long and bus_dma ran out of segments. Defragment the mbuf chain and try again. (Already existed, not changed.) - ENOMEM means bus_dma could not obtain enough bounce buffers at this point in time. Defer sending and try again later. - All other errors, in particular EINVAL, are fatal and prevent the mbuf chain from ever going through. Drop it and report error. - Checking (nsegs == 0) is unnecessary as bus_dmamap_load_mbuf_sg() always reports an error if it is < 1. This prevents broken packets from clogging the interface queue indefinately. Discussed with: scottl Reviewed by: jfv
2006-09-29 13:12:38 +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
if (error) {
Make em(4) handle too many fragmented frame with m_defrag(9). Previously em(4) requeued the failed mbuf chains from bus_dmamap_load_mbuf_sg(9) failure to resend it later. However, bus_dmamap_load_mbuf_sg(9) may never complete its request as the fragmented frames can have more than EM_MAX_SCATTER segments. To handle the above EFBIG case, defragment the frame with m_defrag(9) and free the mbuf chain if it can't deframent the chain due to resource shortage. Reviewed by glebius (with improvements)
2006-08-14 02:21:26 +00:00
adapter->no_tx_dma_setup++;
m_freem(*m_headp);
*m_headp = NULL;
return (error);
}
Handle all error cases from bus_dmamap_load_mbuf_sg(). Those are: - EFBIG means the mbuf chain was too long and bus_dma ran out of segments. Defragment the mbuf chain and try again. (Already existed, not changed.) - ENOMEM means bus_dma could not obtain enough bounce buffers at this point in time. Defer sending and try again later. - All other errors, in particular EINVAL, are fatal and prevent the mbuf chain from ever going through. Drop it and report error. - Checking (nsegs == 0) is unnecessary as bus_dmamap_load_mbuf_sg() always reports an error if it is < 1. This prevents broken packets from clogging the interface queue indefinately. Discussed with: scottl Reviewed by: jfv
2006-09-29 13:12:38 +00:00
} else if (error != 0) {
adapter->no_tx_dma_setup++;
return (error);
Make em(4) handle too many fragmented frame with m_defrag(9). Previously em(4) requeued the failed mbuf chains from bus_dmamap_load_mbuf_sg(9) failure to resend it later. However, bus_dmamap_load_mbuf_sg(9) may never complete its request as the fragmented frames can have more than EM_MAX_SCATTER segments. To handle the above EFBIG case, defragment the frame with m_defrag(9) and free the mbuf chain if it can't deframent the chain due to resource shortage. Reviewed by glebius (with improvements)
2006-08-14 02:21:26 +00:00
}
Prepending an mbuf after loading a DMA map results in unexpected result. So, modify mbuf chains before loading a DMA map.
2006-07-27 00:43:34 +00:00
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
/*
* TSO Hardware workaround, if this packet is not
* TSO, and is only a single descriptor long, and
* it follows a TSO burst, then we need to add a
* sentinel descriptor to prevent premature writeback.
*/
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 ((do_tso == 0) && (txr->tx_tso == TRUE)) {
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
if (nsegs == 1)
tso_desc = TRUE;
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
txr->tx_tso = FALSE;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
}
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +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 (nsegs > (txr->tx_avail - 2)) {
txr->no_desc_avail++;
bus_dmamap_unload(txr->txtag, map);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
return (ENOBUFS);
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
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
m_head = *m_headp;
Prepending an mbuf after loading a DMA map results in unexpected result. So, modify mbuf chains before loading a DMA map.
2006-07-27 00:43:34 +00:00
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
/* Do hardware assists */
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 __FreeBSD_version >= 700000
if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
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 = em_tso_setup(txr, m_head, &txd_upper, &txd_lower);
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 (error != TRUE)
return (ENXIO); /* something foobar */
This delta adds two bug fixes: one that makes HW Offload logic in legacy codepath match the 82575, without this we were seeing bridging fail on 82546 adapters. Secondly, I have limited TSO to PCI Express adapters, I meant to do this and it got dropped in the earlier delta. Next, I am dropping in the latest shared code from our development team, consensus was that this should be done frequently, so I am :) Approved by: pdeuskar
2007-05-16 00:14:23 +00:00
/* we need to make a final sentinel transmit desc */
tso_desc = TRUE;
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
} else
#endif
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 (m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD)
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_transmit_checksum_setup(txr, m_head,
This delta adds two bug fixes: one that makes HW Offload logic in legacy codepath match the 82575, without this we were seeing bridging fail on 82546 adapters. Secondly, I have limited TSO to PCI Express adapters, I meant to do this and it got dropped in the earlier delta. Next, I am dropping in the latest shared code from our development team, consensus was that this should be done frequently, so I am :) Approved by: pdeuskar
2007-05-16 00:14:23 +00:00
&txd_upper, &txd_lower);
Prepending an mbuf after loading a DMA map results in unexpected result. So, modify mbuf chains before loading a DMA map.
2006-07-27 00:43:34 +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
i = txr->next_avail_desc;
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +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
/* Set up our transmit descriptors */
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
for (j = 0; j < nsegs; j++) {
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
bus_size_t seg_len;
bus_addr_t seg_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
tx_buffer = &txr->tx_buffers[i];
ctxd = &txr->tx_base[i];
seg_addr = segs[j].ds_addr;
seg_len = segs[j].ds_len;
/*
** TSO Workaround:
** If this is the last descriptor, we want to
** split it so we have a small final sentinel
*/
if (tso_desc && (j == (nsegs -1)) && (seg_len > 8)) {
seg_len -= 4;
ctxd->buffer_addr = htole64(seg_addr);
ctxd->lower.data = htole32(
adapter->txd_cmd | txd_lower | seg_len);
ctxd->upper.data =
htole32(txd_upper);
if (++i == adapter->num_tx_desc)
i = 0;
/* Now make the sentinel */
++txd_used; /* using an extra txd */
ctxd = &txr->tx_base[i];
tx_buffer = &txr->tx_buffers[i];
ctxd->buffer_addr =
htole64(seg_addr + seg_len);
ctxd->lower.data = htole32(
adapter->txd_cmd | txd_lower | 4);
ctxd->upper.data =
htole32(txd_upper);
last = i;
if (++i == adapter->num_tx_desc)
i = 0;
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
} 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
ctxd->buffer_addr = htole64(seg_addr);
ctxd->lower.data = htole32(
adapter->txd_cmd | txd_lower | seg_len);
ctxd->upper.data =
htole32(txd_upper);
last = i;
if (++i == adapter->num_tx_desc)
i = 0;
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
}
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
tx_buffer->m_head = NULL;
tx_buffer->next_eop = -1;
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
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
txr->next_avail_desc = i;
txr->tx_avail -= nsegs;
if (tso_desc) /* TSO used an extra for sentinel */
txr->tx_avail -= txd_used;
- 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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (m_head->m_flags & M_VLANTAG) {
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 the vlan id. */
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
ctxd->upper.fields.special =
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
htole16(m_head->m_pkthdr.ether_vtag);
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
/* Tell hardware to add tag */
ctxd->lower.data |= htole32(E1000_TXD_CMD_VLE);
}
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
tx_buffer->m_head = m_head;
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
tx_buffer_mapped->map = tx_buffer->map;
Make sure to use the same DMA map in DMA map load/unload operations by remembering a map used in bus_dmamap_load_mbuf_sg(9). I have no idea how it could ever worked before. This fixes a warning generated by a diagnostic check in sun4v iommu driver. Reported by: jb Tested by: jb(sun4v)
2006-07-27 00:26:20 +00:00
tx_buffer->map = map;
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
bus_dmamap_sync(txr->txtag, map, BUS_DMASYNC_PREWRITE);
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
/*
* Last Descriptor of Packet
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
* needs End Of Packet (EOP)
* and Report Status (RS)
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
*/
ctxd->lower.data |=
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
htole32(E1000_TXD_CMD_EOP | E1000_TXD_CMD_RS);
/*
* Keep track in the first buffer which
* descriptor will be written back
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 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
tx_buffer = &txr->tx_buffers[first];
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
tx_buffer->next_eop = last;
- 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
/*
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
* Advance the Transmit Descriptor Tail (TDT), this tells the E1000
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
* that this frame is available to transmit.
*/
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
bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map,
- 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
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
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_TDT(txr->me), i);
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 (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
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_set_promisc(struct adapter *adapter)
- 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
{
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 ifnet *ifp = adapter->ifp;
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;
- 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
if (ifp->if_flags & IFF_PROMISC) {
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);
- 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 if (ifp->if_flags & IFF_ALLMULTI) {
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
}
- 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
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_disable_promisc(struct adapter *adapter)
- 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 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;
- 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
reg_rctl &= (~E1000_RCTL_UPE);
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
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_set_multi(struct adapter *adapter)
- 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
{
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 ifnet *ifp = adapter->ifp;
- 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
struct ifmultiaddr *ifma;
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");
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_82542 &&
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
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
/* Allocate temporary memory to setup array */
mta = malloc(sizeof(u8) *
(ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES),
M_DEVBUF, M_NOWAIT | M_ZERO);
if (mta == NULL)
panic("em_set_multi memory failure\n");
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 __FreeBSD_version < 800000
IF_ADDR_LOCK(ifp);
#else
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_rlock(ifp);
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
#endif
- 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
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
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 (ifma->ifma_addr->sa_family != AF_LINK)
continue;
if (mcnt == MAX_NUM_MULTICAST_ADDRESSES)
break;
bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_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
&mta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
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
mcnt++;
}
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 __FreeBSD_version < 800000
IF_ADDR_UNLOCK(ifp);
#else
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_runlock(ifp);
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
#endif
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
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_82542 &&
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
}
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
free(mta, M_DEVBUF);
- 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
em_local_timer(void *arg)
{
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 = arg;
struct ifnet *ifp = adapter->ifp;
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 tx_ring *txr = adapter->tx_rings;
- 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
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_CORE_LOCK_ASSERT(adapter);
- 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
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_link_status(adapter);
em_update_stats_counters(adapter);
Couple of the fixes needed revising. The ICH8 autoneg was still broken, this change both simplifies the code and plugs a hole where the devise was reset without keeping the management controller at bay :) Second, the 82571 LAA reset problem was incomplete, this addition is necessary. Just one of those days :)
2007-05-31 23:36:21 +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
/* Reset LAA into RAR[0] on 82571 */
if (e1000_get_laa_state_82571(&adapter->hw) == TRUE)
Couple of the fixes needed revising. The ICH8 autoneg was still broken, this change both simplifies the code and plugs a hole where the devise was reset without keeping the management controller at bay :) Second, the 82571 LAA reset problem was incomplete, this addition is necessary. Just one of those days :)
2007-05-31 23:36:21 +00:00
e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 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
if (em_display_debug_stats && ifp->if_drv_flags & IFF_DRV_RUNNING)
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_print_hw_stats(adapter);
Couple of the fixes needed revising. The ICH8 autoneg was still broken, this change both simplifies the code and plugs a hole where the devise was reset without keeping the management controller at bay :) Second, the 82571 LAA reset problem was incomplete, this addition is necessary. Just one of those days :)
2007-05-31 23:36:21 +00:00
Instead of using the legacy if_timer/if_watchdog interface create our own watchdog that piggybacks on the em_local_timer() routine. We suppose that the if_timer/if_watchdog interface should be obsoleted, since it doesn't fit the modern SMP network stack. NIC drivers should create their own watchdogs, that check and clear the timers always holding driver's lock. In collaboration with: jfv, scottl
2006-11-09 16:00:18 +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
** Check for time since any descriptor was cleaned
*/
for (int i = 0; i < adapter->num_queues; i++, txr++) {
EM_TX_LOCK(txr);
if (txr->watchdog_check == FALSE) {
EM_TX_UNLOCK(txr);
continue;
}
if ((ticks - txr->watchdog_time) > EM_WATCHDOG)
goto hung;
EM_TX_UNLOCK(txr);
}
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
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
callout_reset(&adapter->timer, hz, em_local_timer, 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
return;
hung:
device_printf(adapter->dev, "Watchdog timeout -- resetting\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
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
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->watchdog_events++;
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_UNLOCK(txr);
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_init_locked(adapter);
- 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
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_link_status(struct adapter *adapter)
- 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
struct e1000_hw *hw = &adapter->hw;
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 ifnet *ifp = adapter->ifp;
device_t dev = adapter->dev;
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
u32 link_check = 0;
/* 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) {
/* 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);
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
link_check = TRUE;
break;
case e1000_media_type_fiber:
e1000_check_for_link(hw);
link_check = (E1000_READ_REG(hw, E1000_STATUS) &
E1000_STATUS_LU);
break;
case e1000_media_type_internal_serdes:
e1000_check_for_link(hw);
link_check = adapter->hw.mac.serdes_has_link;
break;
default:
case e1000_media_type_unknown:
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
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));
tarc0 &= ~SPEED_MODE_BIT;
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;
ifp->if_baudrate = adapter->link_speed * 1000000;
if_link_state_change(ifp, LINK_STATE_UP);
} else if (!link_check && (adapter->link_active == 1)) {
ifp->if_baudrate = adapter->link_speed = 0;
adapter->link_duplex = 0;
if (bootverbose)
device_printf(dev, "Link is Down\n");
adapter->link_active = 0;
/* Link down, disable watchdog */
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
// JFV change later
//adapter->watchdog_check = FALSE;
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_link_state_change(ifp, LINK_STATE_DOWN);
- 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
}
/*********************************************************************
*
- 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
em_stop(void *arg)
{
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 = arg;
struct ifnet *ifp = adapter->ifp;
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 tx_ring *txr = adapter->tx_rings;
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
EM_CORE_LOCK_ASSERT(adapter);
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");
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
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_disable_intr(adapter);
callout_stop(&adapter->timer);
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
/* Tell the stack that the interface is no longer active */
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
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
/* Unarm watchdog timer. */
for (int i = 0; i < adapter->num_queues; i++, txr++) {
EM_TX_LOCK(txr);
txr->watchdog_check = FALSE;
EM_TX_UNLOCK(txr);
}
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);
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_WUC, 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
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_identify_hardware(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
{
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_t dev = adapter->dev;
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);
A number of small fixes: - duplicate #define in header, thanks to Kevin Lo for pointing out. - incorrect BUSMASTER enable logic, thanks Patrick Oeschger - 82543 fails due to bogus IO BAR logic - Allow 82571 to use MSI interrupts - Checksum Offload for UDP not working on 82575 Approved by:re
2007-09-10 21:50:40 +00:00
if (!((adapter->hw.bus.pci_cmd_word & PCIM_CMD_BUSMASTEREN) &&
(adapter->hw.bus.pci_cmd_word & PCIM_CMD_MEMEN))) {
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, "Memory Access and/or Bus Master bits "
"were not set!\n");
adapter->hw.bus.pci_cmd_word |=
(PCIM_CMD_BUSMASTEREN | PCIM_CMD_MEMEN);
pci_write_config(dev, PCIR_COMMAND,
adapter->hw.bus.pci_cmd_word, 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
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_allocate_pci_resources(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
{
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_t dev = adapter->dev;
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 rid;
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);
}
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->osdep.mem_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->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
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
/* Default to a single queue */
adapter->num_queues = 1;
- 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
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 MSI/X or MSI if PCI Express
*/
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
adapter->msix = em_setup_msix(adapter);
Significant performance improvements for the if_em driver: - Only update the rx ring consumer pointer after running through the rx loop, not with each iteration through the loop. - If possible, use a fast interupt handler instead of an ithread handler. Use the interrupt handler to check and squelch the interrupt, then schedule a taskqueue to do the actual work. This has three benefits: - Eliminates the 'interrupt aliasing' problem found in many chipsets by allowing the driver to mask the interrupt in the NIC instead of the OS masking the interrupt in the APIC. - Allows the driver to control the amount of work done in the interrupt handler. This results in what I call 'adaptive polling', where you get the latency benefits of a quick response to interrupts with the interrupt mitigation and work partitioning of polling. Polling is still an option in the driver, but I consider it orthogonal to this work. - Don't hold the driver lock in the RX handler. The handler and all data associated is effectively serialized already. This eliminates the cost of dropping and reaquiring the lock for every receieved packet. The result is much lower contention for the driver lock, resulting in lower CPU usage and lower latency for interactive workloads. The amount of work done in the taskqueue is controlled by the sysctl dev.em.N.rx_processing_limit and tunable hw.em.rx_process_limit Setting these to -1 effectively removes the limit. The fast interrupt and taskqueue can be disabled by defining NO_EM_FASTINTR. This work has been shown to increase fast-forwarding from ~570 kpps to ~750 kpps (note that the same NIC hardware seems unable to transmit more than 800 kpps, so this increase appears to be limited almost solely by the hardware). Gains have been shown in other workloads, ranging from better performance to elimination of over-saturation livelocks. Thanks to Andre Opperman for his time and resources from his network performance project in performing much of the testing. Thanks to Gleb Smirnoff and Danny Braniss for their help in testing also.
2006-01-11 00:30:25 +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->hw.back = &adapter->osdep;
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
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);
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
}
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 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 Legacy or MSI Interrupt handler
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
*
**********************************************************************/
Revert previous commit and add myself to the list of people who should know better than to commit with a cat in the area.
2009-09-08 13:19:05 +00:00
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_allocate_legacy(struct adapter *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
{
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
device_t dev = adapter->dev;
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, rid = 0;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/* Manually turn off all interrupts */
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);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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 (adapter->msix == 1) /* using MSI */
rid = 1;
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
/* We allocate a single interrupt resource */
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
adapter->res = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE);
if (adapter->res == NULL) {
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
device_printf(dev, "Unable to allocate bus resource: "
"interrupt\n");
return (ENXIO);
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
* Allocate a fast interrupt and the associated
* deferred processing contexts.
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
TASK_INIT(&adapter->que_task, 0, em_handle_que, adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
TASK_INIT(&adapter->link_task, 0, em_handle_link, adapter);
adapter->tq = taskqueue_create_fast("em_taskq", M_NOWAIT,
taskqueue_thread_enqueue, &adapter->tq);
taskqueue_start_threads(&adapter->tq, 1, PI_NET, "%s taskq",
device_get_nameunit(adapter->dev));
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 = bus_setup_intr(dev, adapter->res, INTR_TYPE_NET,
em_irq_fast, NULL, adapter, &adapter->tag)) != 0) {
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
device_printf(dev, "Failed to register fast interrupt "
"handler: %d\n", error);
taskqueue_free(adapter->tq);
adapter->tq = NULL;
return (error);
}
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
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
* This is not really Multiqueue, rather
* its just multiple interrupt vectors.
*
**********************************************************************/
int
em_allocate_msix(struct adapter *adapter)
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
{
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
device_t dev = adapter->dev;
struct tx_ring *txr = adapter->tx_rings;
struct rx_ring *rxr = adapter->rx_rings;
int error, rid, vector = 0;
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
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
/* Make sure all interrupts are disabled */
E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
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 */
for (int i = 0; i < adapter->num_queues; i++, txr++, rxr++) {
/* RX ring */
rid = vector + 1;
rxr->res = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &rid, RF_ACTIVE);
if (rxr->res == NULL) {
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
device_printf(dev,
"Unable to allocate bus resource: "
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
"RX MSIX Interrupt %d\n", i);
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 (ENXIO);
}
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 = bus_setup_intr(dev, rxr->res,
INTR_TYPE_NET | INTR_MPSAFE, NULL, em_msix_rx,
rxr, &rxr->tag)) != 0) {
device_printf(dev, "Failed to register RX handler");
return (error);
}
rxr->msix = vector++; /* NOTE increment vector for TX */
TASK_INIT(&rxr->rx_task, 0, em_handle_rx, rxr);
rxr->tq = taskqueue_create_fast("em_rxq", M_NOWAIT,
taskqueue_thread_enqueue, &rxr->tq);
taskqueue_start_threads(&rxr->tq, 1, PI_NET, "%s rxq",
device_get_nameunit(adapter->dev));
/*
** 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
*/
rxr->ims = 1 << (20 + i);
adapter->ivars |= (8 | rxr->msix) << (i * 4);
/* TX ring */
rid = vector + 1;
txr->res = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &rid, RF_ACTIVE);
if (txr->res == NULL) {
device_printf(dev,
"Unable to allocate bus resource: "
"TX MSIX Interrupt %d\n", i);
return (ENXIO);
}
if ((error = bus_setup_intr(dev, txr->res,
INTR_TYPE_NET | INTR_MPSAFE, NULL, em_msix_tx,
txr, &txr->tag)) != 0) {
device_printf(dev, "Failed to register TX handler");
return (error);
}
txr->msix = vector++; /* Increment vector for next pass */
TASK_INIT(&txr->tx_task, 0, em_handle_tx, txr);
txr->tq = taskqueue_create_fast("em_txq", M_NOWAIT,
taskqueue_thread_enqueue, &txr->tq);
taskqueue_start_threads(&txr->tq, 1, PI_NET, "%s txq",
device_get_nameunit(adapter->dev));
/*
** 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
*/
txr->ims = 1 << (22 + i);
adapter->ivars |= (8 | txr->msix) << (8 + (i * 4));
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
}
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
/* Link interrupt */
++rid;
adapter->res = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &rid, RF_ACTIVE);
if (!adapter->res) {
device_printf(dev,"Unable to allocate "
"bus resource: Link interrupt [%d]\n", rid);
return (ENXIO);
}
/* Set the link handler function */
error = bus_setup_intr(dev, adapter->res,
INTR_TYPE_NET | INTR_MPSAFE, NULL,
em_msix_link, adapter, &adapter->tag);
if (error) {
adapter->res = NULL;
device_printf(dev, "Failed to register LINK handler");
return (error);
}
adapter->linkvec = vector;
adapter->ivars |= (8 | vector) << 16;
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);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
static void
em_free_pci_resources(struct adapter *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
device_t dev = adapter->dev;
struct tx_ring *txr;
struct rx_ring *rxr;
int rid;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
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
** Release all the queue interrupt resources:
*/
for (int i = 0; i < adapter->num_queues; i++) {
txr = &adapter->tx_rings[i];
rxr = &adapter->rx_rings[i];
rid = txr->msix +1;
if (txr->tag != NULL) {
bus_teardown_intr(dev, txr->res, txr->tag);
txr->tag = NULL;
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
if (txr->res != NULL)
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
bus_release_resource(dev, SYS_RES_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
rid, txr->res);
rid = rxr->msix +1;
if (rxr->tag != NULL) {
bus_teardown_intr(dev, rxr->res, rxr->tag);
rxr->tag = NULL;
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
if (rxr->res != NULL)
bus_release_resource(dev, SYS_RES_IRQ,
rid, rxr->res);
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
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 (adapter->linkvec) /* we are doing MSIX */
rid = adapter->linkvec + 1;
else
(adapter->msix != 0) ? (rid = 1):(rid = 0);
if (adapter->tag != NULL) {
bus_teardown_intr(dev, adapter->res, adapter->tag);
adapter->tag = NULL;
}
if (adapter->res != NULL)
bus_release_resource(dev, SYS_RES_IRQ, rid, adapter->res);
if (adapter->msix)
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
pci_release_msi(dev);
- 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
if (adapter->msix_mem != NULL)
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
bus_release_resource(dev, SYS_RES_MEMORY,
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
PCIR_BAR(EM_MSIX_BAR), adapter->msix_mem);
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)
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,
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
PCIR_BAR(0), adapter->memory);
- 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 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)
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,
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_FLASH, adapter->flash);
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 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
* Setup MSI or MSI/X
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 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
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_setup_msix(struct adapter *adapter)
{
device_t dev = adapter->dev;
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
int val = 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
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
/* Setup MSI/X for Hartwell */
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 ((adapter->hw.mac.type == e1000_82574) &&
(em_enable_msix == TRUE)) {
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
/* Map the MSIX BAR */
int rid = PCIR_BAR(EM_MSIX_BAR);
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
adapter->msix_mem = bus_alloc_resource_any(dev,
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
SYS_RES_MEMORY, &rid, RF_ACTIVE);
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 (!adapter->msix_mem) {
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
/* May not be enabled */
device_printf(adapter->dev,
"Unable to map MSIX table \n");
goto msi;
}
val = pci_msix_count(dev);
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 (val != 5) {
bus_release_resource(dev, SYS_RES_MEMORY,
PCIR_BAR(EM_MSIX_BAR), adapter->msix_mem);
adapter->msix_mem = NULL;
device_printf(adapter->dev,
"MSIX vectors wrong, using MSI \n");
goto msi;
}
if (em_msix_queues == 2) {
val = 5;
adapter->num_queues = 2;
} else {
val = 3;
adapter->num_queues = 1;
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
}
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 (pci_alloc_msix(dev, &val) == 0) {
device_printf(adapter->dev,
"Using MSIX interrupts "
"with %d vectors\n", val);
}
return (val);
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:
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
val = pci_msi_count(dev);
if (val == 1 && pci_alloc_msi(dev, &val) == 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
adapter->msix = 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
device_printf(adapter->dev,"Using MSI interrupt\n");
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 (val);
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
}
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
/* Should only happen due to manual invention */
device_printf(adapter->dev,"Setup MSIX failure\n");
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
}
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
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
* Initialize the hardware to a configuration
* as specified by the adapter structure.
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 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
em_reset(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
{
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
device_t dev = adapter->dev;
struct e1000_hw *hw = &adapter->hw;
u16 rx_buffer_size;
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_reset: begin");
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
}
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
* drained a bit. Here we use an arbitary value of 1500 which will
* 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.
*/
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
rx_buffer_size = ((E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10 );
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
hw->fc.high_water = rx_buffer_size -
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
roundup2(adapter->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
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
/* Set Flow control, use the tunable location if sane */
if ((em_fc_setting >= 0) || (em_fc_setting < 4))
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.requested_mode = em_fc_setting;
else
hw->fc.requested_mode = e1000_fc_none;
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
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
/* Override - workaround for PCHLAN issue */
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_pchlan)
hw->fc.requested_mode = e1000_fc_rx_pause;
/* Issue a global reset */
e1000_reset_hw(hw);
E1000_WRITE_REG(hw, E1000_WUC, 0);
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
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
}
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);
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
}
/*********************************************************************
*
* Setup networking device structure and register an interface.
*
**********************************************************************/
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_setup_interface(device_t dev, 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
struct ifnet *ifp;
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
INIT_DEBUGOUT("em_setup_interface: begin");
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
ifp = adapter->ifp = if_alloc(IFT_ETHER);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
if (ifp == NULL)
panic("%s: can not if_alloc()", device_get_nameunit(dev));
Replace the if_name and if_unit members of struct ifnet with new members if_xname, if_dname, and if_dunit. if_xname is the name of the interface and if_dname/unit are the driver name and instance. This change paves the way for interface renaming and enhanced pseudo device creation and configuration symantics. Approved By: re (in principle) Reviewed By: njl, imp Tested On: i386, amd64, sparc64 Obtained From: NetBSD (if_xname)
2003-10-31 18:32:15 +00:00
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
ifp->if_mtu = ETHERMTU;
ifp->if_init = em_init;
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
ifp->if_softc = adapter;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
ifp->if_ioctl = em_ioctl;
ifp->if_start = em_start;
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
IFQ_SET_MAXLEN(&ifp->if_snd, adapter->num_tx_desc - 1);
ifp->if_snd.ifq_drv_maxlen = adapter->num_tx_desc - 1;
Bring in the first chunk of altq driver modifications. This covers the following drivers: bfe(4), em(4), fxp(4), lnc(4), tun(4), de(4) rl(4), sis(4) and xl(4) More patches are pending on: http://peoples.freebsd.org/~mlaier/ Please take a look and tell me if "your" driver is missing, so I can fix this. Tested-by: many No-objection: -current, -net
2004-07-02 12:16:02 +00:00
IFQ_SET_READY(&ifp->if_snd);
- 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
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
ether_ifattach(ifp, adapter->hw.mac.addr);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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
ifp->if_capabilities = ifp->if_capenable = 0;
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
#ifdef EM_MULTIQUEUE
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
/* Multiqueue tx functions */
ifp->if_transmit = em_mq_start;
- 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
ifp->if_qflush = em_qflush;
#endif
- 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
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
ifp->if_capabilities |= IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM;
ifp->if_capenable |= IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM;
/* Enable TSO by default, can disable with ifconfig */
ifp->if_capabilities |= IFCAP_TSO4;
ifp->if_capenable |= IFCAP_TSO4;
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
*/
ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
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
ifp->if_capabilities |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU;
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
ifp->if_capenable |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU;
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
/*
** Dont 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.
*/
ifp->if_capabilities |= IFCAP_VLAN_HWFILTER;
Implemented per-interface polling(4) control.
2004-04-11 21:01:12 +00:00
#ifdef DEVICE_POLLING
ifp->if_capabilities |= IFCAP_POLLING;
#endif
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
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) {
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
ifp->if_capabilities |= IFCAP_WOL;
Change default WOL back to MAGIC only, having multicast enabled causes problems in man environments.
2010-04-28 17:37:30 +00:00
ifp->if_capenable |= IFCAP_WOL_MAGIC;
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
}
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
*/
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
ifmedia_init(&adapter->media, IFM_IMASK,
em_media_change, em_media_status);
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
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
ifmedia_add(&adapter->media, IFM_ETHER | fiber_type | IFM_FDX,
0, NULL);
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
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 {
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
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX,
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
0, NULL);
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
ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX,
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
0, NULL);
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
ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX,
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
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) {
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
ifmedia_add(&adapter->media,
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
IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
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
ifmedia_add(&adapter->media,
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
IFM_ETHER | IFM_1000_T, 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
}
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
ifmedia_add(&adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
ifmedia_set(&adapter->media, IFM_ETHER | IFM_AUTO);
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
/*
* Manage DMA'able memory.
*/
static void
em_dmamap_cb(void *arg, bus_dma_segment_t *segs, int nseg, int error)
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 (error)
return;
*(bus_addr_t *) arg = segs[0].ds_addr;
- 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
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_dma_malloc(struct adapter *adapter, bus_size_t size,
struct em_dma_alloc *dma, int mapflags)
- 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;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(adapter->dev), /* parent */
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
EM_DBA_ALIGN, 0, /* alignment, bounds */
- 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
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
size, /* maxsize */
1, /* nsegments */
size, /* maxsegsize */
0, /* flags */
NULL, /* lockfunc */
NULL, /* lockarg */
&dma->dma_tag);
if (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
device_printf(adapter->dev,
"%s: bus_dma_tag_create failed: %d\n",
- 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
__func__, error);
goto fail_0;
}
error = bus_dmamem_alloc(dma->dma_tag, (void**) &dma->dma_vaddr,
Add COHERENT to descriptor mem allocation for the benefit of ARM (request from Olivier Houchard), its a noop on most architectures and goodness on those that use it.
2007-11-28 23:04:11 +00:00
BUS_DMA_NOWAIT | BUS_DMA_COHERENT, &dma->dma_map);
- 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) {
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(adapter->dev,
"%s: bus_dmamem_alloc(%ju) failed: %d\n",
- 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
__func__, (uintmax_t)size, error);
goto fail_2;
}
- 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
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
dma->dma_paddr = 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
error = bus_dmamap_load(dma->dma_tag, dma->dma_map, dma->dma_vaddr,
size, em_dmamap_cb, &dma->dma_paddr, mapflags | BUS_DMA_NOWAIT);
if (error || dma->dma_paddr == 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(adapter->dev,
"%s: bus_dmamap_load failed: %d\n",
- 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
__func__, error);
goto fail_3;
}
return (0);
- 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
fail_3:
- 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
bus_dmamap_unload(dma->dma_tag, dma->dma_map);
- 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
fail_2:
- 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
bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map);
bus_dma_tag_destroy(dma->dma_tag);
- 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
fail_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
dma->dma_map = NULL;
dma->dma_tag = NULL;
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
}
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_dma_free(struct adapter *adapter, struct em_dma_alloc *dma)
- 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
{
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
if (dma->dma_tag == NULL)
return;
if (dma->dma_map != NULL) {
bus_dmamap_sync(dma->dma_tag, dma->dma_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(dma->dma_tag, dma->dma_map);
bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map);
dma->dma_map = 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
bus_dma_tag_destroy(dma->dma_tag);
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
dma->dma_tag = NULL;
- 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
/*********************************************************************
*
* Allocate memory for the transmit and receive rings, and then
* the descriptors associated with each, called only once at attach.
*
**********************************************************************/
static int
em_allocate_queues(struct adapter *adapter)
{
device_t dev = adapter->dev;
struct tx_ring *txr = NULL;
struct rx_ring *rxr = NULL;
int rsize, tsize, error = E1000_SUCCESS;
int txconf = 0, rxconf = 0;
/* Allocate the TX ring struct memory */
if (!(adapter->tx_rings =
(struct tx_ring *) malloc(sizeof(struct tx_ring) *
adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
device_printf(dev, "Unable to allocate TX ring memory\n");
error = ENOMEM;
goto fail;
}
/* Now allocate the RX */
if (!(adapter->rx_rings =
(struct rx_ring *) malloc(sizeof(struct rx_ring) *
adapter->num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
device_printf(dev, "Unable to allocate RX ring memory\n");
error = ENOMEM;
goto rx_fail;
}
tsize = roundup2(adapter->num_tx_desc *
sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
/*
* Now set up the TX queues, txconf is needed to handle the
* possibility that things fail midcourse and we need to
* undo memory gracefully
*/
for (int i = 0; i < adapter->num_queues; i++, txconf++) {
/* Set up some basics */
txr = &adapter->tx_rings[i];
txr->adapter = adapter;
txr->me = i;
/* Initialize the TX lock */
snprintf(txr->mtx_name, sizeof(txr->mtx_name), "%s:tx(%d)",
device_get_nameunit(dev), txr->me);
mtx_init(&txr->tx_mtx, txr->mtx_name, NULL, MTX_DEF);
if (em_dma_malloc(adapter, tsize,
&txr->txdma, BUS_DMA_NOWAIT)) {
device_printf(dev,
"Unable to allocate TX Descriptor memory\n");
error = ENOMEM;
goto err_tx_desc;
}
txr->tx_base = (struct e1000_tx_desc *)txr->txdma.dma_vaddr;
bzero((void *)txr->tx_base, tsize);
if (em_allocate_transmit_buffers(txr)) {
device_printf(dev,
"Critical Failure setting up transmit buffers\n");
error = ENOMEM;
goto err_tx_desc;
}
#if __FreeBSD_version >= 800000
/* Allocate a buf ring */
txr->br = buf_ring_alloc(4096, M_DEVBUF,
M_WAITOK, &txr->tx_mtx);
#endif
}
/*
* Next the RX queues...
*/
rsize = roundup2(adapter->num_rx_desc *
sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
for (int i = 0; i < adapter->num_queues; i++, rxconf++) {
rxr = &adapter->rx_rings[i];
rxr->adapter = adapter;
rxr->me = i;
/* Initialize the RX lock */
snprintf(rxr->mtx_name, sizeof(rxr->mtx_name), "%s:rx(%d)",
device_get_nameunit(dev), txr->me);
mtx_init(&rxr->rx_mtx, rxr->mtx_name, NULL, MTX_DEF);
if (em_dma_malloc(adapter, rsize,
&rxr->rxdma, BUS_DMA_NOWAIT)) {
device_printf(dev,
"Unable to allocate RxDescriptor memory\n");
error = ENOMEM;
goto err_rx_desc;
}
rxr->rx_base = (struct e1000_rx_desc *)rxr->rxdma.dma_vaddr;
bzero((void *)rxr->rx_base, rsize);
/* Allocate receive buffers for the ring*/
if (em_allocate_receive_buffers(rxr)) {
device_printf(dev,
"Critical Failure setting up receive buffers\n");
error = ENOMEM;
goto err_rx_desc;
}
}
return (0);
err_rx_desc:
for (rxr = adapter->rx_rings; rxconf > 0; rxr++, rxconf--)
em_dma_free(adapter, &rxr->rxdma);
err_tx_desc:
for (txr = adapter->tx_rings; txconf > 0; txr++, txconf--)
em_dma_free(adapter, &txr->txdma);
free(adapter->rx_rings, M_DEVBUF);
rx_fail:
Small changes preparing for MFC, need to conditionalize the buf_ring_free call, and lem is missing the WOL change put into em.
2010-05-14 22:18:34 +00:00
#if __FreeBSD_version >= 800000
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
buf_ring_free(txr->br, M_DEVBUF);
Small changes preparing for MFC, need to conditionalize the buf_ring_free call, and lem is missing the WOL change put into em.
2010-05-14 22:18:34 +00:00
#endif
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
free(adapter->tx_rings, M_DEVBUF);
fail:
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
/*********************************************************************
*
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
* Allocate memory for tx_buffer structures. The tx_buffer stores all
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
* the information needed to transmit a packet on the wire. This is
* called only once at attach, setup is done every reset.
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
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_allocate_transmit_buffers(struct tx_ring *txr)
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
struct adapter *adapter = txr->adapter;
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_t dev = adapter->dev;
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 em_buffer *txbuf;
int error, i;
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
- 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
* Setup DMA descriptor areas.
- 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
if ((error = bus_dma_tag_create(bus_get_dma_tag(dev),
1, 0, /* alignment, bounds */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
EM_TSO_SIZE, /* maxsize */
EM_MAX_SCATTER, /* nsegments */
PAGE_SIZE, /* maxsegsize */
0, /* flags */
NULL, /* lockfunc */
NULL, /* lockfuncarg */
&txr->txtag))) {
device_printf(dev,"Unable to allocate TX DMA tag\n");
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
goto fail;
- 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
}
- 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
if (!(txr->tx_buffers =
(struct em_buffer *) malloc(sizeof(struct em_buffer) *
adapter->num_tx_desc, M_DEVBUF, M_NOWAIT | M_ZERO))) {
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
device_printf(dev, "Unable to allocate tx_buffer memory\n");
error = ENOMEM;
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
goto fail;
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
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
/* Create the descriptor buffer dma maps */
txbuf = txr->tx_buffers;
for (i = 0; i < adapter->num_tx_desc; i++, txbuf++) {
error = bus_dmamap_create(txr->txtag, 0, &txbuf->map);
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
if (error != 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, "Unable to create TX DMA map\n");
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
goto fail;
}
}
- 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
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
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
/* We free all, it handles case where we are in the middle */
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_free_transmit_structures(adapter);
return (error);
}
/*********************************************************************
*
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
* Initialize a transmit ring.
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
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_setup_transmit_ring(struct tx_ring *txr)
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
{
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 adapter *adapter = txr->adapter;
struct em_buffer *txbuf;
int i;
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
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
/* Clear the old descriptor contents */
EM_TX_LOCK(txr);
bzero((void *)txr->tx_base,
(sizeof(struct e1000_tx_desc)) * adapter->num_tx_desc);
/* Reset indices */
txr->next_avail_desc = 0;
txr->next_to_clean = 0;
/* Free any existing tx buffers. */
txbuf = txr->tx_buffers;
for (i = 0; i < adapter->num_tx_desc; i++, txbuf++) {
if (txbuf->m_head != NULL) {
bus_dmamap_sync(txr->txtag, txbuf->map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(txr->txtag, txbuf->map);
m_freem(txbuf->m_head);
txbuf->m_head = NULL;
}
/* clear the watch index */
txbuf->next_eop = -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
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 number of descriptors available */
txr->tx_avail = adapter->num_tx_desc;
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 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
bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map,
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_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
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_UNLOCK(txr);
}
/*********************************************************************
*
* Initialize all transmit rings.
*
**********************************************************************/
static void
em_setup_transmit_structures(struct adapter *adapter)
{
struct tx_ring *txr = adapter->tx_rings;
for (int i = 0; i < adapter->num_queues; i++, txr++)
em_setup_transmit_ring(txr);
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 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;
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
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_initialize_transmit_unit(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
{
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 tx_ring *txr = adapter->tx_rings;
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 tctl, 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
- 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
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
for (int i = 0; i < adapter->num_queues; i++, txr++) {
u64 bus_addr = txr->txdma.dma_paddr;
/* Base and Len of TX Ring */
E1000_WRITE_REG(hw, E1000_TDLEN(i),
adapter->num_tx_desc * sizeof(struct e1000_tx_desc));
E1000_WRITE_REG(hw, E1000_TDBAH(i),
(u32)(bus_addr >> 32));
E1000_WRITE_REG(hw, E1000_TDBAL(i),
(u32)bus_addr);
/* 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)));
txr->watchdog_check = FALSE;
}
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_82542:
tipg = DEFAULT_82542_TIPG_IPGT;
tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
tipg |= DEFAULT_82542_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
break;
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
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;
- 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
}
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_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
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_82540)
E1000_WRITE_REG(&adapter->hw, E1000_TADV,
adapter->tx_abs_int_delay.value);
if ((adapter->hw.mac.type == e1000_82571) ||
(adapter->hw.mac.type == e1000_82572)) {
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
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(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
tarc |= SPEED_MODE_BIT;
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
E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
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
} else if (adapter->hw.mac.type == e1000_80003es2lan) {
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
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(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
tarc |= 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
E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
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
tarc |= 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
E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
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
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
adapter->txd_cmd = E1000_TXD_CMD_IFCS;
if (adapter->tx_int_delay.value > 0)
adapter->txd_cmd |= E1000_TXD_CMD_IDE;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Program the Transmit Control Register */
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
tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
tctl &= ~E1000_TCTL_CT;
A few small but significant fixes: - Coverity Prevent(tm) CID 1906 a bogus use of bzero where unneeded. - ICH8 systems autoneg to 100 rather than 1000, this can also be seen in 82573, the logic was backwards. - On new 82575 quadports half duplex tx speed is slow... this was due to overwriting TCTL reg rather than adding bits.
2007-05-30 23:32:21 +00:00
tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
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)
tctl |= E1000_TCTL_MULR;
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
/* This write will effectively turn on the transmit unit. */
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_TCTL, tctl);
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
}
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
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 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
* Free all transmit rings.
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_free_transmit_structures(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
{
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 tx_ring *txr = adapter->tx_rings;
- 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
for (int i = 0; i < adapter->num_queues; i++, txr++) {
EM_TX_LOCK(txr);
em_free_transmit_buffers(txr);
em_dma_free(adapter, &txr->txdma);
EM_TX_UNLOCK(txr);
EM_TX_LOCK_DESTROY(txr);
}
- 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
free(adapter->tx_rings, M_DEVBUF);
}
/*********************************************************************
*
* Free transmit ring related data structures.
*
**********************************************************************/
static void
em_free_transmit_buffers(struct tx_ring *txr)
{
struct adapter *adapter = txr->adapter;
struct em_buffer *txbuf;
INIT_DEBUGOUT("free_transmit_ring: begin");
if (txr->tx_buffers == NULL)
return;
for (int i = 0; i < adapter->num_tx_desc; i++) {
txbuf = &txr->tx_buffers[i];
if (txbuf->m_head != NULL) {
bus_dmamap_sync(txr->txtag, txbuf->map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(txr->txtag,
txbuf->map);
m_freem(txbuf->m_head);
txbuf->m_head = NULL;
if (txbuf->map != NULL) {
bus_dmamap_destroy(txr->txtag,
txbuf->map);
txbuf->map = 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
}
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
} else if (txbuf->map != NULL) {
bus_dmamap_unload(txr->txtag,
txbuf->map);
bus_dmamap_destroy(txr->txtag,
txbuf->map);
txbuf->map = 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
}
}
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 __FreeBSD_version >= 800000
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 (txr->br != NULL)
buf_ring_free(txr->br, M_DEVBUF);
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
#endif
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 (txr->tx_buffers != NULL) {
free(txr->tx_buffers, M_DEVBUF);
txr->tx_buffers = NULL;
}
if (txr->txtag != NULL) {
bus_dma_tag_destroy(txr->txtag);
txr->txtag = NULL;
}
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
}
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
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
/*********************************************************************
*
* The offload context needs to be set when we transfer the first
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
* packet of a particular protocol (TCP/UDP). This routine has been
* enhanced to deal with inserted VLAN headers, and IPV6 (not complete)
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
*
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
* Added back the old method of keeping the current context type
* and not setting if unnecessary, as this is reported to be a
* big performance win. -jfv
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
**********************************************************************/
Back out TX/RX descriptor/buffer management changes from earier commit. We are having panics with the driver under stress with jumbo frames. Unfortunately we didnot catch it during our regular test cycle. I am going to MFC the backout immediately.
2002-08-26 18:59:16 +00:00
static void
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_transmit_checksum_setup(struct tx_ring *txr, struct mbuf *mp,
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 *txd_upper, u32 *txd_lower)
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 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 adapter *adapter = txr->adapter;
struct e1000_context_desc *TXD = NULL;
- 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
struct em_buffer *tx_buffer;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
struct ether_vlan_header *eh;
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
struct ip *ip = NULL;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
struct ip6_hdr *ip6;
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 cur, ehdrlen;
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 cmd, hdr_len, ip_hlen;
u16 etype;
u8 ipproto;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
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
cmd = hdr_len = ipproto = 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
cur = txr->next_avail_desc;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*
* Determine where frame payload starts.
* Jump over vlan headers if already present,
* helpful for QinQ too.
*/
eh = mtod(mp, struct ether_vlan_header *);
if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
etype = ntohs(eh->evl_proto);
ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
} else {
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
etype = ntohs(eh->evl_encap_proto);
ehdrlen = ETHER_HDR_LEN;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*
* We only support TCP/UDP for IPv4 and IPv6 for the moment.
* TODO: Support SCTP too when it hits the tree.
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
*/
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
switch (etype) {
case ETHERTYPE_IP:
ip = (struct ip *)(mp->m_data + ehdrlen);
ip_hlen = ip->ip_hl << 2;
/* Setup of IP header checksum. */
if (mp->m_pkthdr.csum_flags & CSUM_IP) {
/*
* Start offset for header checksum calculation.
* End offset for header checksum calculation.
* Offset of place to put the checksum.
*/
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
TXD = (struct e1000_context_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
&txr->tx_base[cur];
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
TXD->lower_setup.ip_fields.ipcss = ehdrlen;
TXD->lower_setup.ip_fields.ipcse =
htole16(ehdrlen + ip_hlen);
TXD->lower_setup.ip_fields.ipcso =
ehdrlen + offsetof(struct ip, ip_sum);
cmd |= E1000_TXD_CMD_IP;
*txd_upper |= E1000_TXD_POPTS_IXSM << 8;
}
Change em_transmit_checksum_setup() to deal with already inserted vlan headers, IP options and add skeleton IPv6 support. The new code structure can also be easily enhanced to support new/more protocols (SCTP) in the future. Reviewed by: jfv
2006-09-29 13:37:26 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (mp->m_len < ehdrlen + ip_hlen)
return; /* failure */
Change em_transmit_checksum_setup() to deal with already inserted vlan headers, IP options and add skeleton IPv6 support. The new code structure can also be easily enhanced to support new/more protocols (SCTP) in the future. Reviewed by: jfv
2006-09-29 13:37:26 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
hdr_len = ehdrlen + ip_hlen;
ipproto = ip->ip_p;
break;
case ETHERTYPE_IPV6:
ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen);
ip_hlen = sizeof(struct ip6_hdr); /* XXX: No header stacking. */
if (mp->m_len < ehdrlen + ip_hlen)
return; /* failure */
/* IPv6 doesn't have a header checksum. */
hdr_len = ehdrlen + ip_hlen;
ipproto = ip6->ip6_nxt;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
break;
default:
*txd_upper = 0;
*txd_lower = 0;
return;
Change em_transmit_checksum_setup() to deal with already inserted vlan headers, IP options and add skeleton IPv6 support. The new code structure can also be easily enhanced to support new/more protocols (SCTP) in the future. Reviewed by: jfv
2006-09-29 13:37:26 +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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
switch (ipproto) {
case IPPROTO_TCP:
if (mp->m_pkthdr.csum_flags & CSUM_TCP) {
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
*txd_lower = E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
*txd_upper |= E1000_TXD_POPTS_TXSM << 8;
/* no need for context if already set */
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 (txr->last_hw_offload == CSUM_TCP)
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
return;
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
txr->last_hw_offload = CSUM_TCP;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*
* Start offset for payload checksum calculation.
* End offset for payload checksum calculation.
* Offset of place to put the checksum.
*/
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
TXD = (struct e1000_context_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
&txr->tx_base[cur];
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
TXD->upper_setup.tcp_fields.tucss = hdr_len;
TXD->upper_setup.tcp_fields.tucse = htole16(0);
TXD->upper_setup.tcp_fields.tucso =
hdr_len + offsetof(struct tcphdr, th_sum);
cmd |= E1000_TXD_CMD_TCP;
}
break;
case IPPROTO_UDP:
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
if (mp->m_pkthdr.csum_flags & CSUM_UDP) {
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
*txd_lower = E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
*txd_upper |= E1000_TXD_POPTS_TXSM << 8;
/* no need for context if already set */
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 (txr->last_hw_offload == CSUM_UDP)
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
return;
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
txr->last_hw_offload = CSUM_UDP;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*
* Start offset for header checksum calculation.
* End offset for header checksum calculation.
* Offset of place to put the checksum.
*/
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
TXD = (struct e1000_context_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
&txr->tx_base[cur];
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
TXD->upper_setup.tcp_fields.tucss = hdr_len;
TXD->upper_setup.tcp_fields.tucse = htole16(0);
TXD->upper_setup.tcp_fields.tucso =
hdr_len + offsetof(struct udphdr, uh_sum);
}
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
/* Fall Thru */
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
default:
break;
}
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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
TXD->tcp_seg_setup.data = htole32(0);
TXD->cmd_and_length =
htole32(adapter->txd_cmd | E1000_TXD_CMD_DEXT | cmd);
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
tx_buffer = &txr->tx_buffers[cur];
- 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
tx_buffer->m_head = NULL;
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
tx_buffer->next_eop = -1;
- 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
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 (++cur == adapter->num_tx_desc)
cur = 0;
Back out TX/RX descriptor/buffer management changes from earier commit. We are having panics with the driver under stress with jumbo frames. Unfortunately we didnot catch it during our regular test cycle. I am going to MFC the backout immediately.
2002-08-26 18:59:16 +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
txr->tx_avail--;
txr->next_avail_desc = cur;
- 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
}
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
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +00:00
/**********************************************************************
*
* Setup work for hardware segmentation offload (TSO)
*
**********************************************************************/
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 bool
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_tso_setup(struct tx_ring *txr, struct mbuf *mp, u32 *txd_upper,
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 *txd_lower)
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +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
struct adapter *adapter = txr->adapter;
struct e1000_context_desc *TXD;
struct em_buffer *tx_buffer;
struct ether_vlan_header *eh;
struct ip *ip;
struct ip6_hdr *ip6;
struct tcphdr *th;
int cur, ehdrlen, hdr_len, ip_hlen, isip6;
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 etype;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*
* This function could/should be extended to support IP/IPv6
* fragmentation as well. But as they say, one step at a time.
*/
/*
* Determine where frame payload starts.
* Jump over vlan headers if already present,
* helpful for QinQ too.
*/
eh = mtod(mp, struct ether_vlan_header *);
if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
etype = ntohs(eh->evl_proto);
ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
} else {
etype = ntohs(eh->evl_encap_proto);
ehdrlen = ETHER_HDR_LEN;
}
/* Ensure we have at least the IP+TCP header in the first mbuf. */
if (mp->m_len < ehdrlen + sizeof(struct ip) + sizeof(struct tcphdr))
return FALSE; /* -1 */
/*
* We only support TCP for IPv4 and IPv6 (notyet) for the moment.
* TODO: Support SCTP too when it hits the tree.
*/
switch (etype) {
case ETHERTYPE_IP:
isip6 = 0;
ip = (struct ip *)(mp->m_data + ehdrlen);
if (ip->ip_p != IPPROTO_TCP)
return FALSE; /* 0 */
ip->ip_len = 0;
ip->ip_sum = 0;
ip_hlen = ip->ip_hl << 2;
if (mp->m_len < ehdrlen + ip_hlen + sizeof(struct tcphdr))
return FALSE; /* -1 */
th = (struct tcphdr *)((caddr_t)ip + ip_hlen);
#if 1
th->th_sum = in_pseudo(ip->ip_src.s_addr,
ip->ip_dst.s_addr, htons(IPPROTO_TCP));
#else
th->th_sum = mp->m_pkthdr.csum_data;
#endif
break;
case ETHERTYPE_IPV6:
isip6 = 1;
return FALSE; /* Not supported yet. */
ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen);
if (ip6->ip6_nxt != IPPROTO_TCP)
return FALSE; /* 0 */
ip6->ip6_plen = 0;
ip_hlen = sizeof(struct ip6_hdr); /* XXX: no header stacking. */
if (mp->m_len < ehdrlen + ip_hlen + sizeof(struct tcphdr))
return FALSE; /* -1 */
th = (struct tcphdr *)((caddr_t)ip6 + ip_hlen);
#if 0
th->th_sum = in6_pseudo(ip6->ip6_src, ip->ip6_dst,
htons(IPPROTO_TCP)); /* XXX: function notyet. */
#else
th->th_sum = mp->m_pkthdr.csum_data;
#endif
break;
default:
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
return FALSE;
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
hdr_len = ehdrlen + ip_hlen + (th->th_off << 2);
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
*txd_lower = (E1000_TXD_CMD_DEXT | /* Extended descr type */
E1000_TXD_DTYP_D | /* Data descr type */
E1000_TXD_CMD_TSE); /* Do TSE on this packet */
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/* IP and/or TCP header checksum calculation and insertion. */
*txd_upper = ((isip6 ? 0 : E1000_TXD_POPTS_IXSM) |
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
E1000_TXD_POPTS_TXSM) << 8;
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
cur = txr->next_avail_desc;
tx_buffer = &txr->tx_buffers[cur];
TXD = (struct e1000_context_desc *) &txr->tx_base[cur];
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/* IPv6 doesn't have a header checksum. */
if (!isip6) {
/*
* Start offset for header checksum calculation.
* End offset for header checksum calculation.
* Offset of place put the checksum.
*/
TXD->lower_setup.ip_fields.ipcss = ehdrlen;
TXD->lower_setup.ip_fields.ipcse =
htole16(ehdrlen + ip_hlen - 1);
TXD->lower_setup.ip_fields.ipcso =
ehdrlen + offsetof(struct ip, ip_sum);
}
/*
* Start offset for payload checksum calculation.
* End offset for payload checksum calculation.
* Offset of place to put the checksum.
*/
TXD->upper_setup.tcp_fields.tucss =
ehdrlen + ip_hlen;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
TXD->upper_setup.tcp_fields.tucse = 0;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
TXD->upper_setup.tcp_fields.tucso =
ehdrlen + ip_hlen + offsetof(struct tcphdr, th_sum);
/*
* Payload size per packet w/o any headers.
* Length of all headers up to payload.
*/
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
TXD->tcp_seg_setup.fields.mss = htole16(mp->m_pkthdr.tso_segsz);
TXD->tcp_seg_setup.fields.hdr_len = hdr_len;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
TXD->cmd_and_length = htole32(adapter->txd_cmd |
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
E1000_TXD_CMD_DEXT | /* Extended descr */
E1000_TXD_CMD_TSE | /* TSE 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
(isip6 ? 0 : E1000_TXD_CMD_IP) |
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
E1000_TXD_CMD_TCP | /* Do TCP checksum */
(mp->m_pkthdr.len - (hdr_len))); /* Total len */
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +00:00
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
tx_buffer->m_head = 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
tx_buffer->next_eop = -1;
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +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 (++cur == adapter->num_tx_desc)
cur = 0;
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +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
txr->tx_avail--;
txr->next_avail_desc = cur;
txr->tx_tso = TRUE;
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +00:00
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
return TRUE;
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +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
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
- 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
/**********************************************************************
*
* Examine each tx_buffer in the used queue. If the hardware is done
* processing the packet then free associated resources. The
- 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
* tx_buffer is put back on the free queue.
- 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
static bool
em_txeof(struct tx_ring *txr)
- 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
struct adapter *adapter = txr->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
int first, last, done, num_avail;
struct em_buffer *tx_buffer;
struct e1000_tx_desc *tx_desc, *eop_desc;
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 ifnet *ifp = adapter->ifp;
- 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
EM_TX_LOCK_ASSERT(txr);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +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 (txr->tx_avail == adapter->num_tx_desc)
return (FALSE);
- 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
num_avail = txr->tx_avail;
first = txr->next_to_clean;
tx_desc = &txr->tx_base[first];
tx_buffer = &txr->tx_buffers[first];
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
last = tx_buffer->next_eop;
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
eop_desc = &txr->tx_base[last];
- 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
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +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
* What this does is get the index of the
* first descriptor AFTER the EOP of the
* first packet, that way we can do the
* simple comparison on the inner while loop.
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
*/
if (++last == adapter->num_tx_desc)
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 = 0;
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
done = last;
- 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
bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map,
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_DMASYNC_POSTREAD);
while (eop_desc->upper.fields.status & E1000_TXD_STAT_DD) {
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
/* We clean the range of the packet */
while (first != done) {
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
tx_desc->upper.data = 0;
tx_desc->lower.data = 0;
tx_desc->buffer_addr = 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
++num_avail;
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
if (tx_buffer->m_head) {
ifp->if_opackets++;
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
bus_dmamap_sync(txr->txtag,
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
tx_buffer->map,
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
BUS_DMASYNC_POSTWRITE);
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
bus_dmamap_unload(txr->txtag,
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
tx_buffer->map);
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
m_freem(tx_buffer->m_head);
tx_buffer->m_head = NULL;
}
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
tx_buffer->next_eop = -1;
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
txr->watchdog_time = ticks;
- 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
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 (++first == adapter->num_tx_desc)
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
first = 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
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
tx_buffer = &txr->tx_buffers[first];
tx_desc = &txr->tx_base[first];
- 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
}
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
/* See if we can continue to the next packet */
last = tx_buffer->next_eop;
if (last != -1) {
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
eop_desc = &txr->tx_base[last];
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
/* Get new done point */
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 (++last == adapter->num_tx_desc) last = 0;
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
done = last;
} else
break;
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
bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map,
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_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
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
txr->next_to_clean = first;
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
/*
Thanks for Michael Tuexen for tracking down a path where the watchdog timer was not being rearmed in txeof, and also a missing case in the new code. MFC after: 2 weeks
2009-04-27 17:22:14 +00:00
* If we have enough room, clear IFF_DRV_OACTIVE to
* tell the stack that it is OK to send packets.
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 there are no pending descriptors, clear the watchdog.
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 (num_avail > EM_TX_CLEANUP_THRESHOLD) {
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
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 (num_avail == adapter->num_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
txr->watchdog_check = FALSE;
txr->tx_avail = num_avail;
return (FALSE);
Thanks for Michael Tuexen for tracking down a path where the watchdog timer was not being rearmed in txeof, and also a missing case in the new code. MFC after: 2 weeks
2009-04-27 17:22:14 +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
}
Thanks for Michael Tuexen for tracking down a path where the watchdog timer was not being rearmed in txeof, and also a missing case in the new code. MFC after: 2 weeks
2009-04-27 17:22:14 +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
txr->tx_avail = num_avail;
return (TRUE);
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 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 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
* Refresh RX descriptor mbufs from system mbuf buffer pool.
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 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
em_refresh_mbufs(struct rx_ring *rxr, int limit)
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 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 adapter *adapter = rxr->adapter;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
struct mbuf *m;
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
bus_dma_segment_t segs[1];
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
bus_dmamap_t map;
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 em_buffer *rxbuf;
int i, error, nsegs, cleaned;
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
i = rxr->next_to_refresh;
cleaned = -1;
while (i != limit) {
m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
if (m == NULL)
goto update;
m->m_len = m->m_pkthdr.len = MCLBYTES;
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
if (adapter->max_frame_size <= (MCLBYTES - ETHER_ALIGN))
m_adj(m, ETHER_ALIGN);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
/*
* Using memory from the mbuf cluster pool, invoke the
* bus_dma machinery to arrange the memory mapping.
*/
error = bus_dmamap_load_mbuf_sg(rxr->rxtag, rxr->rx_sparemap,
m, segs, &nsegs, BUS_DMA_NOWAIT);
if (error != 0) {
m_free(m);
goto update;
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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 nsegs is wrong then the stack is corrupt. */
KASSERT(nsegs == 1, ("Too many segments returned!"));
rxbuf = &rxr->rx_buffers[i];
if (rxbuf->m_head != NULL)
bus_dmamap_unload(rxr->rxtag, rxbuf->map);
map = rxbuf->map;
rxbuf->map = rxr->rx_sparemap;
rxr->rx_sparemap = map;
bus_dmamap_sync(rxr->rxtag,
rxbuf->map, BUS_DMASYNC_PREREAD);
rxbuf->m_head = m;
rxr->rx_base[i].buffer_addr = htole64(segs[0].ds_addr);
cleaned = i;
/* Calculate next index */
if (++i == adapter->num_rx_desc)
i = 0;
/* This is the work marker for refresh */
rxr->next_to_refresh = i;
}
update:
Three changes: - add CRC stripping to the RX side, this was handled by some obscure code in rxeof previously, its easier to simply have the hardware strip it now. - Add back an ALTQ change that slipped between the cracks - Add an update to the watchdog_time in the xmit code, not doing this in ixgbe caused problems, think its needed here as well.
2010-04-08 19:13:42 +00:00
bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
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 (cleaned != -1) /* Update tail index */
E1000_WRITE_REG(&adapter->hw,
E1000_RDT(rxr->me), cleaned);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
}
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
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
* Allocate memory for rx_buffer structures. Since we use one
* rx_buffer per received packet, the maximum number of rx_buffer's
* that we'll need is equal to the number of receive descriptors
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
* that we've allocated.
*
**********************************************************************/
static int
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_allocate_receive_buffers(struct rx_ring *rxr)
- 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
struct adapter *adapter = rxr->adapter;
device_t dev = adapter->dev;
struct em_buffer *rxbuf;
int 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
rxr->rx_buffers = malloc(sizeof(struct em_buffer) *
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->num_rx_desc, M_DEVBUF, M_NOWAIT | M_ZERO);
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 (rxr->rx_buffers == 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 rx_buffer memory\n");
return (ENOMEM);
}
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
error = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
- 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
1, 0, /* alignment, bounds */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
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
MCLBYTES, /* maxsize */
- 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
1, /* nsegments */
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
MCLBYTES, /* maxsegsize */
Nuke invalid use of BUS_DMA_ALLOCNOW.
2006-07-27 00:29:44 +00:00
0, /* flags */
- 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
NULL, /* lockfunc */
NULL, /* lockarg */
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
&rxr->rxtag);
- 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) {
device_printf(dev, "%s: bus_dma_tag_create failed %d\n",
__func__, error);
goto fail;
}
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
/* Create the spare map (used by getbuf) */
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 = bus_dmamap_create(rxr->rxtag, BUS_DMA_NOWAIT,
&rxr->rx_sparemap);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (error) {
device_printf(dev, "%s: bus_dmamap_create failed: %d\n",
__func__, error);
goto fail;
}
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
rxbuf = rxr->rx_buffers;
for (int i = 0; i < adapter->num_rx_desc; i++, rxbuf++) {
rxbuf = &rxr->rx_buffers[i];
error = bus_dmamap_create(rxr->rxtag, BUS_DMA_NOWAIT,
&rxbuf->map);
- 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) {
device_printf(dev, "%s: bus_dmamap_create failed: %d\n",
__func__, error);
goto fail;
}
}
return (0);
- 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
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
fail:
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_free_receive_structures(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
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
}
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
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 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
* Initialize a receive ring and its buffers.
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
**********************************************************************/
static int
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_setup_receive_ring(struct rx_ring *rxr)
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 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 adapter *adapter = rxr->adapter;
struct em_buffer *rxbuf;
bus_dma_segment_t seg[1];
int rsize, nsegs, error;
/* Clear the ring contents */
EM_RX_LOCK(rxr);
rsize = roundup2(adapter->num_rx_desc *
sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
bzero((void *)rxr->rx_base, rsize);
/*
** Free current RX buffer structs and their mbufs
*/
for (int i = 0; i < adapter->num_rx_desc; i++) {
rxbuf = &rxr->rx_buffers[i];
if (rxbuf->m_head != NULL) {
bus_dmamap_sync(rxr->rxtag, rxbuf->map,
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
BUS_DMASYNC_POSTREAD);
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
bus_dmamap_unload(rxr->rxtag, rxbuf->map);
m_freem(rxbuf->m_head);
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
}
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
}
/* Now replenish the mbufs */
for (int j = 0; j != adapter->num_rx_desc; ++j) {
rxbuf = &rxr->rx_buffers[j];
rxbuf->m_head = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
if (rxbuf->m_head == NULL)
panic("RX ring hdr initialization failed!\n");
rxbuf->m_head->m_len = MCLBYTES;
rxbuf->m_head->m_flags &= ~M_HASFCS; /* we strip it */
rxbuf->m_head->m_pkthdr.len = MCLBYTES;
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
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
/* Get the memory mapping */
error = bus_dmamap_load_mbuf_sg(rxr->rxtag,
rxbuf->map, rxbuf->m_head, seg,
&nsegs, BUS_DMA_NOWAIT);
if (error != 0)
panic("RX ring dma initialization failed!\n");
bus_dmamap_sync(rxr->rxtag,
rxbuf->map, BUS_DMASYNC_PREREAD);
/* Update descriptor */
rxr->rx_base[j].buffer_addr = htole64(seg[0].ds_addr);
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
}
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
/* Setup our descriptor indices */
rxr->next_to_check = 0;
rxr->next_to_refresh = 0;
bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map,
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
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
- 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
EM_RX_UNLOCK(rxr);
return (0);
}
/*********************************************************************
*
* Initialize all receive rings.
*
**********************************************************************/
static int
em_setup_receive_structures(struct adapter *adapter)
{
struct rx_ring *rxr = adapter->rx_rings;
int j;
for (j = 0; j < adapter->num_queues; j++, rxr++)
if (em_setup_receive_ring(rxr))
goto fail;
- 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);
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
fail:
/*
* Free RX buffers allocated so far, we will only handle
* the rings that completed, the failing case will have
* cleaned up for itself. 'j' failed, so its the terminus.
*/
for (int i = 0; i < j; ++i) {
rxr = &adapter->rx_rings[i];
for (int n = 0; n < adapter->num_rx_desc; n++) {
struct em_buffer *rxbuf;
rxbuf = &rxr->rx_buffers[n];
if (rxbuf->m_head != NULL) {
bus_dmamap_sync(rxr->rxtag, rxbuf->map,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(rxr->rxtag, rxbuf->map);
m_freem(rxbuf->m_head);
rxbuf->m_head = NULL;
}
}
}
return (ENOBUFS);
}
/*********************************************************************
*
* Free all receive rings.
*
**********************************************************************/
static void
em_free_receive_structures(struct adapter *adapter)
{
struct rx_ring *rxr = adapter->rx_rings;
for (int i = 0; i < adapter->num_queues; i++, rxr++) {
em_free_receive_buffers(rxr);
/* Free the ring memory as well */
em_dma_free(adapter, &rxr->rxdma);
EM_RX_LOCK_DESTROY(rxr);
}
free(adapter->rx_rings, M_DEVBUF);
}
/*********************************************************************
*
* Free receive ring data structures
*
**********************************************************************/
static void
em_free_receive_buffers(struct rx_ring *rxr)
{
struct adapter *adapter = rxr->adapter;
struct em_buffer *rxbuf = NULL;
INIT_DEBUGOUT("free_receive_buffers: begin");
if (rxr->rx_sparemap) {
bus_dmamap_destroy(rxr->rxtag, rxr->rx_sparemap);
rxr->rx_sparemap = NULL;
}
if (rxr->rx_buffers != NULL) {
for (int i = 0; i < adapter->num_rx_desc; i++) {
rxbuf = &rxr->rx_buffers[i];
if (rxbuf->map != NULL) {
bus_dmamap_sync(rxr->rxtag, rxbuf->map,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(rxr->rxtag, rxbuf->map);
bus_dmamap_destroy(rxr->rxtag, rxbuf->map);
}
if (rxbuf->m_head != NULL) {
m_freem(rxbuf->m_head);
rxbuf->m_head = NULL;
}
}
free(rxr->rx_buffers, M_DEVBUF);
rxr->rx_buffers = NULL;
}
if (rxr->rxtag != NULL) {
bus_dma_tag_destroy(rxr->rxtag);
rxr->rxtag = NULL;
}
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
}
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
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
#define MAX_INTS_PER_SEC 8000
#define DEFAULT_ITR 1000000000/(MAX_INTS_PER_SEC * 256)
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_initialize_receive_unit(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
{
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 rx_ring *rxr = adapter->rx_rings;
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 ifnet *ifp = adapter->ifp;
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;
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
u64 bus_addr;
u32 rctl, rxcsum;
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);
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
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_RADV,
adapter->rx_abs_int_delay.value);
/*
* 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);
- 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
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
/*
** When using MSIX interrupts we need to throttle
** using the EITR register (82574 only)
*/
A few more changes from yongari: - code flow in handler could let interrupt be reenabled when not wanted. - change where the RX lock is taken to improve performance. - adapter->msix is true for MSI systems also, it needs to explicitly test for 82574, good one :)
2010-04-09 23:15:37 +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);
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
/* Disable accelerated ackknowledge */
if (adapter->hw.mac.type == e1000_82574)
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_RFCTL, E1000_RFCTL_ACK_DIS);
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
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 (ifp->if_capenable & IFCAP_RXCSUM) {
rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
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
rxcsum |= (E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
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_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
}
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
/*
** 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
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
for (int i = 0; i < adapter->num_queues; i++, rxr++) {
/* Setup the Base and Length of the Rx Descriptor Ring */
bus_addr = rxr->rxdma.dma_paddr;
E1000_WRITE_REG(hw, E1000_RDLEN(i),
adapter->num_rx_desc * sizeof(struct e1000_rx_desc));
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 */
E1000_WRITE_REG(hw, E1000_RDH(i), 0);
E1000_WRITE_REG(hw, E1000_RDT(i), adapter->num_rx_desc - 1);
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
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 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);
Three changes: - add CRC stripping to the RX side, this was handled by some obscure code in rxeof previously, its easier to simply have the hardware strip it now. - Add back an ALTQ change that slipped between the cracks - Add an update to the watchdog_time in the xmit code, not doing this in ixgbe caused problems, think its needed here as well.
2010-04-08 19:13:42 +00:00
/* Strip the CRC */
rctl |= E1000_RCTL_SECRC;
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
/* Make sure VLAN Filters are off */
rctl &= ~E1000_RCTL_VFE;
rctl &= ~E1000_RCTL_SBP;
rctl |= E1000_RCTL_SZ_2048;
if (ifp->if_mtu > ETHERMTU)
rctl |= E1000_RCTL_LPE;
else
rctl &= ~E1000_RCTL_LPE;
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
/* 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
}
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
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 routine executes in interrupt context. It replenishes
* the mbufs in the descriptor and sends data which has been
* dma'ed into host memory to upper layer.
*
- 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
* We loop at most count times if count is > 0, or until done if
* count < 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
*
* For polling we also now return the number of cleaned packets
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
*********************************************************************/
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
static bool
em_rxeof(struct rx_ring *rxr, int count, int *done)
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 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 adapter *adapter = rxr->adapter;
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
struct ifnet *ifp = adapter->ifp;
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 mbuf *mp, *sendmp;
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
u8 status = 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
u16 len;
int i, processed, rxdone = 0;
bool eop;
struct e1000_rx_desc *cur;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Address the LOD that some are seeing, put the RX lock back in rxeof (I could see little point in taking it out), and now release it before the stack entry. Also, make it so the 82574 does not configure for multiqueue when its not used in the stack.
2010-04-28 19:22:52 +00:00
EM_RX_LOCK(rxr);
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
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
for (i = rxr->next_to_check, processed = 0; count != 0;) {
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
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
break;
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
bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
cur = &rxr->rx_base[i];
status = cur->status;
mp = sendmp = NULL;
if ((status & E1000_RXD_STAT_DD) == 0)
break;
len = le16toh(cur->length);
eop = (status & E1000_RXD_STAT_EOP) != 0;
count--;
if ((cur->errors & E1000_RXD_ERR_FRAME_ERR_MASK) == 0) {
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Assign correct length to the current fragment */
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
mp = rxr->rx_buffers[i].m_head;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
mp->m_len = len;
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 (rxr->fmp == NULL) {
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
mp->m_pkthdr.len = len;
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
rxr->fmp = mp; /* Store the first mbuf */
rxr->lmp = mp;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
} else {
/* Chain mbuf's together */
mp->m_flags &= ~M_PKTHDR;
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
rxr->lmp->m_next = mp;
rxr->lmp = rxr->lmp->m_next;
rxr->fmp->m_pkthdr.len += len;
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
if (eop) {
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
rxr->fmp->m_pkthdr.rcvif = ifp;
Cleanup from __FreeBSD_version.
2005-10-14 10:34:46 +00:00
ifp->if_ipackets++;
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_receive_checksum(cur, rxr->fmp);
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
#ifndef __NO_STRICT_ALIGNMENT
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->max_frame_size >
Apply alignment fixup only when programmed frame size is greater than MCLBYTES - ETHER_ALIGN. Previously it applied the alignment fixup code for oversized frames which would result in reduced performance on strict alignment archs.
2006-08-14 00:36:53 +00:00
(MCLBYTES - ETHER_ALIGN) &&
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_fixup_rx(rxr) != 0)
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
goto skip;
#endif
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (status & E1000_RXD_STAT_VP) {
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
rxr->fmp->m_pkthdr.ether_vtag =
(le16toh(cur->special) &
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
E1000_RXD_SPC_VLAN_MASK);
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
rxr->fmp->m_flags |= M_VLANTAG;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
}
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
#ifdef EM_MULTIQUEUE
rxr->fmp->m_pkthdr.flowid = curcpu;
rxr->fmp->m_flags |= M_FLOWID;
#endif
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
#ifndef __NO_STRICT_ALIGNMENT
skip:
#endif
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
sendmp = rxr->fmp;
rxr->fmp = NULL;
rxr->lmp = NULL;
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
} else {
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
ifp->if_ierrors++;
/* Reuse loaded DMA map and just update mbuf chain */
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
mp = rxr->rx_buffers[i].m_head;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
mp->m_len = mp->m_pkthdr.len = MCLBYTES;
mp->m_data = mp->m_ext.ext_buf;
mp->m_next = 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
if (adapter->max_frame_size <=
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
(MCLBYTES - ETHER_ALIGN))
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
m_adj(mp, ETHER_ALIGN);
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 (rxr->fmp != NULL) {
m_freem(rxr->fmp);
rxr->fmp = NULL;
rxr->lmp = NULL;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
sendmp = NULL;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
Don't update Rx descriptor status in two different functions. Suggested by: pdeuskar Reviewed by: pdeuskar
2006-08-16 23:55:34 +00:00
/* Zero out the receive descriptors status. */
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
cur->status = 0;
++rxdone; /* cumulative for POLL */
++processed;
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
/* Advance our pointers to the next descriptor. */
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 (++i == adapter->num_rx_desc)
From the PR: The receive function em_process_receive_interrupts() unlocks the adapter while ether_input() processes the packet, and then locks it back. In the meantime, em_init() may be called, either from em_watchdog() from softclock interrupt or from the ifconfig(8) program. The em_init() resets the card, in particular it sets adapter->next_rx_desc_to_check to 0 and resets hardware RX Head and Tail descriptor pointers. The loop in em_process_receive_interrupts() does not expect these things to change, and a mess may result. This fixes long wedges of em(4) interfaces receive part under high load and IP fastforwarding enabled. PR: kern/87418 Submitted by: Dmitrij Tejblum <tejblum yandex-team.ru>
2005-10-14 11:00:15 +00:00
i = 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
/* Send to the stack */
Address the LOD that some are seeing, put the RX lock back in rxeof (I could see little point in taking it out), and now release it before the stack entry. Also, make it so the 82574 does not configure for multiqueue when its not used in the stack.
2010-04-28 19:22:52 +00:00
if (sendmp != NULL) {
rxr->next_to_check = i;
EM_RX_UNLOCK(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
(*ifp->if_input)(ifp, sendmp);
Address the LOD that some are seeing, put the RX lock back in rxeof (I could see little point in taking it out), and now release it before the stack entry. Also, make it so the 82574 does not configure for multiqueue when its not used in the stack.
2010-04-28 19:22:52 +00:00
EM_RX_LOCK(rxr);
i = rxr->next_to_check;
}
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
/* Only refresh mbufs every 8 descriptors */
if (processed == 8) {
em_refresh_mbufs(rxr, i);
processed = 0;
From the PR: The receive function em_process_receive_interrupts() unlocks the adapter while ether_input() processes the packet, and then locks it back. In the meantime, em_init() may be called, either from em_watchdog() from softclock interrupt or from the ifconfig(8) program. The em_init() resets the card, in particular it sets adapter->next_rx_desc_to_check to 0 and resets hardware RX Head and Tail descriptor pointers. The loop in em_process_receive_interrupts() does not expect these things to change, and a mess may result. This fixes long wedges of em(4) interfaces receive part under high load and IP fastforwarding enabled. PR: kern/87418 Submitted by: Dmitrij Tejblum <tejblum yandex-team.ru>
2005-10-14 11:00:15 +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
}
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
/* Catch any remaining refresh work */
if (processed != 0) {
em_refresh_mbufs(rxr, i);
processed = 0;
}
rxr->next_to_check = i;
Changes from John Baldwin adding to last commit, change rxeof api for poll friendliness, and eliminate unnecessary link tasklet use. Thanks John!
2010-06-16 16:37:36 +00:00
if (done != NULL)
*done = rxdone;
Address the LOD that some are seeing, put the RX lock back in rxeof (I could see little point in taking it out), and now release it before the stack entry. Also, make it so the 82574 does not configure for multiqueue when its not used in the stack.
2010-04-28 19:22:52 +00:00
EM_RX_UNLOCK(rxr);
Incorporate suggested improvements from yongari. Also, from feedback, make the multiqueue code an option (EM_MULTIQUEUE) that is off by default. Problems have been seen with UDP when its on.
2010-04-09 18:42:15 +00:00
return ((status & E1000_RXD_STAT_DD) ? TRUE : FALSE);
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 jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
#ifndef __NO_STRICT_ALIGNMENT
/*
* When jumbo frames are enabled we should realign entire payload on
* architecures with strict alignment. This is serious design mistake of 8254x
* as it nullifies DMA operations. 8254x just allows RX buffer size to be
* 2048/4096/8192/16384. What we really want is 2048 - ETHER_ALIGN to align its
* payload. On architecures without strict alignment restrictions 8254x still
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
* performs unaligned memory access which would reduce the performance too.
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
* To avoid copying over an entire frame to align, we allocate a new mbuf and
* copy ethernet header to the new mbuf. The new mbuf is prepended into the
* existing mbuf chain.
*
* Be aware, best performance of the 8254x is achived only when jumbo frame is
* not used at all on architectures with strict alignment.
*/
static int
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_fixup_rx(struct rx_ring *rxr)
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +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
struct adapter *adapter = rxr->adapter;
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
struct mbuf *m, *n;
int error;
error = 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
m = rxr->fmp;
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
if (m->m_len <= (MCLBYTES - ETHER_HDR_LEN)) {
bcopy(m->m_data, m->m_data + ETHER_HDR_LEN, m->m_len);
m->m_data += ETHER_HDR_LEN;
} else {
MGETHDR(n, M_DONTWAIT, MT_DATA);
if (n != NULL) {
bcopy(m->m_data, n->m_data, ETHER_HDR_LEN);
m->m_data += ETHER_HDR_LEN;
m->m_len -= ETHER_HDR_LEN;
n->m_len = ETHER_HDR_LEN;
M_MOVE_PKTHDR(n, m);
n->m_next = m;
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
rxr->fmp = n;
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +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->dropped_pkts++;
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
m_freem(rxr->fmp);
rxr->fmp = 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
error = ENOMEM;
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
}
}
return (error);
}
#endif
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
* Verify that the hardware indicated that the checksum is valid.
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
* Inform the stack about the status of checksum so that stack
* doesn't spend time verifying the checksum.
*
*********************************************************************/
static void
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_receive_checksum(struct e1000_rx_desc *rx_desc, struct mbuf *mp)
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 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
/* Ignore Checksum bit is set */
if (rx_desc->status & E1000_RXD_STAT_IXSM) {
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
mp->m_pkthdr.csum_flags = 0;
return;
}
if (rx_desc->status & E1000_RXD_STAT_IPCS) {
/* Did it pass? */
if (!(rx_desc->errors & E1000_RXD_ERR_IPE)) {
/* IP Checksum Good */
mp->m_pkthdr.csum_flags = CSUM_IP_CHECKED;
mp->m_pkthdr.csum_flags |= CSUM_IP_VALID;
} else {
mp->m_pkthdr.csum_flags = 0;
}
}
if (rx_desc->status & E1000_RXD_STAT_TCPCS) {
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
/* Did it pass? */
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
if (!(rx_desc->errors & E1000_RXD_ERR_TCPE)) {
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
mp->m_pkthdr.csum_flags |=
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
mp->m_pkthdr.csum_data = htons(0xffff);
}
}
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 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
* This routine is run via an vlan
* config EVENT
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
*/
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
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
em_register_vlan(void *arg, struct ifnet *ifp, 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
{
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
struct adapter *adapter = ifp->if_softc;
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
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
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
if (ifp->if_softc != arg) /* Not our event */
This delta fixes two bugs: - When a vlan event occurs a check was not made that the event was actually for the interface, thus resulting in a panic. All three drivers have this vulnerability. Add a check for this condition. - Secondly, there was a duplicate buf_ring free in the em driver resulting in a panic on unload. Remove. Approved by: re
2009-07-24 16:57:49 +00:00
return;
if ((vtag == 0) || (vtag > 4095)) /* Invalid ID */
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
return;
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;
em_shadow_vfta[index] |= (1 << bit);
++adapter->num_vlans;
/* Re-init to load the changes */
em_init(adapter);
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
}
/*
* This routine is run via an vlan
* unconfig EVENT
*/
static void
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
em_unregister_vlan(void *arg, struct ifnet *ifp, 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
{
struct adapter *adapter = ifp->if_softc;
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
u32 index, bit;
Improvement on the last change, this gives a precise way to tell the one and only interface that a vlan event is for. Thanks to John Baldwin for the patch. Approved by: re
2009-07-24 21:35:52 +00:00
if (ifp->if_softc != arg)
This delta fixes two bugs: - When a vlan event occurs a check was not made that the event was actually for the interface, thus resulting in a panic. All three drivers have this vulnerability. Add a check for this condition. - Secondly, there was a duplicate buf_ring free in the em driver resulting in a panic on unload. Remove. Approved by: re
2009-07-24 16:57:49 +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
if ((vtag == 0) || (vtag > 4095)) /* Invalid */
return;
index = (vtag >> 5) & 0x7F;
bit = vtag & 0x1F;
em_shadow_vfta[index] &= ~(1 << bit);
--adapter->num_vlans;
/* Re-init to load the changes */
em_init(adapter);
}
static void
em_setup_vlan_hw_support(struct adapter *adapter)
{
struct e1000_hw *hw = &adapter->hw;
u32 reg;
/*
** 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.
*/
if (adapter->num_vlans == 0)
return;
/*
** A soft reset zero's out the VFTA, so
** we need to repopulate it now.
*/
for (int i = 0; i < EM_VFTA_SIZE; i++)
if (em_shadow_vfta[i] != 0)
E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
i, em_shadow_vfta[i]);
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);
/* Update the frame size */
E1000_WRITE_REG(&adapter->hw, E1000_RLPML,
adapter->max_frame_size + VLAN_TAG_SIZE);
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
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_enable_intr(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
{
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) {
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, EM_EIAC, EM_MSIX_MASK);
ims_mask |= EM_MSIX_MASK;
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
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
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_disable_intr(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
{
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;
A few more changes from yongari: - code flow in handler could let interrupt be reenabled when not wanted. - change where the RX lock is taken to improve performance. - adapter->msix is true for MSI systems also, it needs to explicitly test for 82574, good one :)
2010-04-09 23:15:37 +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
E1000_WRITE_REG(hw, EM_EIAC, 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
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
* to disable special hardware management features
*/
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);
/* 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;
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);
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
}
/*
* 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
em_get_wakeup(device_t dev)
{
struct adapter *adapter = device_get_softc(dev);
u16 eeprom_data = 0, device_id, apme_mask;
adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
apme_mask = EM_EEPROM_APME;
switch (adapter->hw.mac.type) {
case e1000_82573:
case e1000_82583:
adapter->has_amt = TRUE;
/* Falls thru */
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:
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);
/*
* 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
*/
device_id = pci_get_device(dev);
switch (device_id) {
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:
/* if quad port adapter, disable WoL on all but port A */
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;
break;
}
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
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_enable_wakeup(device_t dev)
{
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
struct adapter *adapter = device_get_softc(dev);
struct ifnet *ifp = adapter->ifp;
u32 pmc, ctrl, ctrl_ext, rctl;
u16 status;
if ((pci_find_extcap(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
/* 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);
E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
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_disable_gig_wol_ich8lan(&adapter->hw);
e1000_hv_phy_powerdown_workaround_ich8lan(&adapter->hw);
}
/* 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);
}
/*
** Determine type of Wakeup: note that wol
** is set with all bits on by default.
*/
if ((ifp->if_capenable & IFCAP_WOL_MAGIC) == 0)
adapter->wol &= ~E1000_WUFC_MAG;
if ((ifp->if_capenable & IFCAP_WOL_MCAST) == 0)
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);
}
if (adapter->hw.mac.type == e1000_pchlan) {
if (em_enable_phy_wakeup(adapter))
return;
} else {
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);
/* Request PME */
status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
if (ifp->if_capenable & IFCAP_WOL)
status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 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
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
/*
** WOL in the newer chipset interfaces (pchlan)
** require thing to be copied into the phy
*/
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 */
for (int i = 0; i < adapter->hw.mac.rar_entry_count; i++) {
mreg = E1000_READ_REG(hw, E1000_RAL(i));
e1000_write_phy_reg(hw, BM_RAR_L(i), (u16)(mreg & 0xFFFF));
e1000_write_phy_reg(hw, BM_RAR_M(i),
(u16)((mreg >> 16) & 0xFFFF));
mreg = E1000_READ_REG(hw, E1000_RAH(i));
e1000_write_phy_reg(hw, BM_RAR_H(i), (u16)(mreg & 0xFFFF));
e1000_write_phy_reg(hw, BM_RAR_CTRL(i),
(u16)((mreg >> 16) & 0xFFFF));
}
/* 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,
E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN);
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
em_led_func(void *arg, int onoff)
{
struct adapter *adapter = arg;
EM_CORE_LOCK(adapter);
if (onoff) {
e1000_setup_led(&adapter->hw);
e1000_led_on(&adapter->hw);
} else {
e1000_led_off(&adapter->hw);
e1000_cleanup_led(&adapter->hw);
}
EM_CORE_UNLOCK(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
/**********************************************************************
*
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
struct ifnet *ifp;
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);
adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
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 */
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.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCH);
adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCH);
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);
if (adapter->hw.mac.type >= e1000_82543) {
adapter->stats.algnerrc +=
E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
adapter->stats.rxerrc +=
E1000_READ_REG(&adapter->hw, E1000_RXERRC);
adapter->stats.tncrs +=
E1000_READ_REG(&adapter->hw, E1000_TNCRS);
adapter->stats.cexterr +=
E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
adapter->stats.tsctc +=
E1000_READ_REG(&adapter->hw, E1000_TSCTC);
adapter->stats.tsctfc +=
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
}
ifp = adapter->ifp;
ifp->if_collisions = adapter->stats.colc;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Rx Errors */
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
ifp->if_ierrors = adapter->dropped_pkts + adapter->stats.rxerrc +
adapter->stats.crcerrs + adapter->stats.algnerrc +
adapter->stats.ruc + adapter->stats.roc +
Overhaul Rx path to recover from mbuf cluster allocation failure. o Create one more spare DMA map for Rx handler to recover from bus_dmamap_load_mbuf_sg(9) failure. o Make sure to update status bit in Rx descriptors even if we failed to allocate a new buffer. Previously it resulted in stuck condition and em_handle_rxtx task took up all available CPU cycles. o Don't blindly unload DMA map. Reuse loaded DMA map if received packet has errors. This would speed up Rx processing a bit under heavy load as it does not need to reload DMA map in case of error. (bus_dmamap_load_mbuf_sg(9) is the most expensive call in driver context.) o Update if_iqdrops counter if it can't allocate a mbuf cluster. With this change it's now possible to see queue dropped packets with netstat(1). o Update mbuf_cluster_failed counter if fixup code failed to allocate mbuf header. o Return ENOBUFS instead of ENOMEM in case of Rx fixup failure. o Make adapter->lmp NULL in case of Rx fixup failure. Strictly specking it's not necessary for correct operation but it makes the intention clear. o Remove now unused dropped_pkts member in softc. With these changes em(4) should survive mbuf cluster allocation failure on Rx path. Reviewed by: pdeuskar, glebius (with improvements)
2006-08-14 01:50:54 +00:00
adapter->stats.mpc + adapter->stats.cexterr;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Tx Errors */
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
ifp->if_oerrors = adapter->stats.ecol +
adapter->stats.latecol + adapter->watchdog_events;
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 routine is called only when em_display_debug_stats is enabled.
* This routine provides a way to take a look at important statistics
* maintained by the driver and hardware.
*
**********************************************************************/
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_print_debug_info(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
{
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_t dev = adapter->dev;
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
u8 *hw_addr = adapter->hw.hw_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
struct rx_ring *rxr = adapter->rx_rings;
struct tx_ring *txr = adapter->tx_rings;
- 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, "Adapter hardware address = %p \n", hw_addr);
device_printf(dev, "CTRL = 0x%x RCTL = 0x%x \n",
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_CTRL),
E1000_READ_REG(&adapter->hw, E1000_RCTL));
- 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, "Packet buffer = Tx=%dk Rx=%dk \n",
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_PBA) & 0xffff0000) >> 16),\
(E1000_READ_REG(&adapter->hw, E1000_PBA) & 0xffff) );
- 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, "Flow control watermarks high = %d low = %d\n",
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.fc.high_water,
adapter->hw.fc.low_water);
- 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, "tx_int_delay = %d, tx_abs_int_delay = %d\n",
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_TIDV),
E1000_READ_REG(&adapter->hw, E1000_TADV));
- 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, "rx_int_delay = %d, rx_abs_int_delay = %d\n",
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_RDTR),
E1000_READ_REG(&adapter->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
for (int i = 0; i < adapter->num_queues; i++, txr++) {
device_printf(dev, "Queue(%d) tdh = %d, tdt = %d\n", i,
E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
Fix lint build problem.
2010-03-30 17:05:24 +00:00
device_printf(dev, "TX(%d) no descriptors avail event = %ld\n",
txr->me, txr->no_desc_avail);
device_printf(dev, "TX(%d) MSIX IRQ Handled = %ld\n",
txr->me, txr->tx_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
device_printf(dev, "Num Tx descriptors avail = %d\n",
txr->tx_avail);
device_printf(dev, "Tx Descriptors not avail1 = %ld\n",
txr->no_desc_avail);
}
for (int i = 0; i < adapter->num_queues; i++, rxr++) {
Fix lint build problem.
2010-03-30 17:05:24 +00:00
device_printf(dev, "RX(%d) MSIX IRQ Handled = %ld\n",
rxr->me, rxr->rx_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
device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
E1000_READ_REG(&adapter->hw, E1000_RDH(i)),
E1000_READ_REG(&adapter->hw, E1000_RDT(i)));
}
- 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, "Std mbuf failed = %ld\n",
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->mbuf_alloc_failed);
- 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, "Std mbuf cluster failed = %ld\n",
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->mbuf_cluster_failed);
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, "Driver dropped packets = %ld\n",
adapter->dropped_pkts);
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
}
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
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_print_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
{
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_t dev = adapter->dev;
- 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, "Excessive collisions = %lld\n",
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
(long long)adapter->stats.ecol);
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 (DEBUG_HW > 0) /* Dont output these errors normally */
- 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, "Symbol errors = %lld\n",
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
(long long)adapter->stats.symerrs);
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
#endif
- 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, "Sequence errors = %lld\n",
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
(long long)adapter->stats.sec);
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, "Defer count = %lld\n",
(long long)adapter->stats.dc);
device_printf(dev, "Missed Packets = %lld\n",
(long long)adapter->stats.mpc);
- 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, "Receive No Buffers = %lld\n",
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
(long long)adapter->stats.rnbc);
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
/* RLEC is inaccurate on some hardware, calculate our own. */
device_printf(dev, "Receive Length Errors = %lld\n",
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
((long long)adapter->stats.roc + (long long)adapter->stats.ruc));
- 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, "Receive errors = %lld\n",
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
(long long)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
device_printf(dev, "Crc errors = %lld\n",
(long long)adapter->stats.crcerrs);
- 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, "Alignment errors = %lld\n",
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
(long long)adapter->stats.algnerrc);
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
device_printf(dev, "Collision/Carrier extension errors = %lld\n",
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
(long long)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
device_printf(dev, "watchdog timeouts = %ld\n",
adapter->watchdog_events);
device_printf(dev, "XON Rcvd = %lld\n",
(long long)adapter->stats.xonrxc);
device_printf(dev, "XON Xmtd = %lld\n",
(long long)adapter->stats.xontxc);
device_printf(dev, "XOFF Rcvd = %lld\n",
(long long)adapter->stats.xoffrxc);
device_printf(dev, "XOFF Xmtd = %lld\n",
(long long)adapter->stats.xofftxc);
- 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, "Good Packets Rcvd = %lld\n",
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
(long long)adapter->stats.gprc);
- 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, "Good Packets Xmtd = %lld\n",
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
(long long)adapter->stats.gptc);
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
device_printf(dev, "TSO Contexts Xmtd = %lld\n",
(long long)adapter->stats.tsctc);
device_printf(dev, "TSO Contexts Failed = %lld\n",
(long long)adapter->stats.tsctfc);
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.
*
**********************************************************************/
static void
em_print_nvm_info(struct adapter *adapter)
{
u16 eeprom_data;
int i, j, row = 0;
/* 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);
}
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
static int
em_sysctl_debug_info(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
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
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
- 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 (result == 1) {
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 = (struct adapter *)arg1;
em_print_debug_info(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
}
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.
*/
if (result == 2) {
adapter = (struct adapter *)arg1;
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
}
static int
em_sysctl_stats(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
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
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
- 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 (result == 1) {
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 = (struct adapter *)arg1;
em_print_hw_stats(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
}
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
}
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
info = (struct em_int_delay_info *)arg1;
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);
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;
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
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_CORE_LOCK(adapter);
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);
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_CORE_UNLOCK(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
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
static void
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_add_rx_process_limit(struct adapter *adapter, const char *name,
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
const char *description, int *limit, int value)
{
*limit = value;
SYSCTL_ADD_INT(device_get_sysctl_ctx(adapter->dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW, limit, value, description);
}
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
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
Reference in New Issue Copy Permalink