d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
3120b9d428
freebsd-skq/sys/dev/e1000/if_em.c

5393 lines
155 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
/******************************************************************************
Copyright (c) 2001-2008, Intel Corporation
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"
#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>
#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>
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
#ifdef EM_TIMESYNC
#include <sys/ioccom.h>
#include <sys/time.h>
#endif
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>
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
*********************************************************************/
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
char em_driver_version[] = "6.9.6";
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_82540EM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82540EM_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82540EP, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82540EP_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82540EP_LP, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82541EI, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82541ER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82541ER_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82541EI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82541GI, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82541GI_LF, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82541GI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
{ 0x8086, E1000_DEV_ID_82542, 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_82543GC_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82543GC_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82544EI_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82544EI_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82544GC_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82544GC_LOM, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82545EM_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82545EM_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82545GM_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82545GM_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82545GM_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82546EB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82546EB_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82546GB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82546GB_FIBER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82546GB_SERDES, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82546GB_PCIE, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, 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_82546GB_QUAD_COPPER_KSP3,
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_82547EI, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82547EI_MOBILE, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82547GI, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 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},
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},
- 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
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},
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},
- 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 *);
static void em_start_locked(struct ifnet *ifp);
static int em_ioctl(struct ifnet *, u_long, caddr_t);
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
static void em_watchdog(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_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 *);
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_allocate_legacy(struct adapter *adapter);
static int em_allocate_msix(struct adapter *adapter);
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 *);
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_hardware_init(struct adapter *);
static void em_setup_interface(device_t, 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_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 *);
static int em_setup_receive_structures(struct adapter *);
static void em_initialize_receive_unit(struct adapter *);
static void em_enable_intr(struct adapter *);
static void em_disable_intr(struct adapter *);
static void em_free_transmit_structures(struct adapter *);
static void em_free_receive_structures(struct adapter *);
static void em_update_stats_counters(struct adapter *);
static void em_txeof(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_tx_purge(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 int em_allocate_receive_structures(struct adapter *);
static int em_allocate_transmit_structures(struct adapter *);
static int em_rxeof(struct adapter *, 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
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_fixup_rx(struct adapter *);
Add jumbo frame support for architectures with strict alignment. Reviewed by: glebius
2005-12-16 08:29:43 +00:00
#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 void em_receive_checksum(struct adapter *, struct e1000_rx_desc *,
- 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 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 void em_transmit_checksum_setup(struct adapter *, 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 *);
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
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 bool em_tso_setup(struct adapter *, struct mbuf *,
u32 *, u32 *);
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 /* FreeBSD_version >= 700000 */
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 *);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
static int em_get_buf(struct adapter *, 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);
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_xmit(struct adapter *, 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 void em_smartspeed(struct adapter *);
static int em_82547_fifo_workaround(struct adapter *, int);
static void em_82547_update_fifo_head(struct adapter *, int);
static int em_82547_tx_fifo_reset(struct 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
static void em_82547_move_tail(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 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);
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 u32 em_fill_descriptors (bus_addr_t address, u32 length,
- 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
PDESC_ARRAY desc_array);
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 *);
static void em_enable_wakeup(device_t);
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 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
#ifdef EM_TIMESYNC
/* Precision Time sync support */
static int em_tsync_init(struct adapter *);
static void em_tsync_disable(struct adapter *);
#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
#ifdef EM_LEGACY_IRQ
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_intr(void *);
#else /* FAST IRQ */
#if __FreeBSD_version < 700000
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_irq_fast(void *);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
#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
static int em_irq_fast(void *);
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
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 *);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_add_rx_process_limit(struct adapter *, const char *,
const char *, int *, int);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
static void em_handle_rxtx(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
static void em_handle_rx(void *context, int pending);
static void em_handle_tx(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);
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
#endif /* EM_LEGACY_IRQ */
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
};
static devclass_t em_devclass;
- Express hard dependencies on bus (pci, isa, pccard) and network layer (ether). - Don't abuse module names to facilitate ifconfig module loading; such abuse isn't really needed. (And if we do need type information associated with a module then we should make it explicit and not use hacks.)
2003-04-15 06:37:30 +00:00
DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
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);
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);
- 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
static int em_rxd = EM_DEFAULT_RXD;
static int em_txd = EM_DEFAULT_TXD;
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
static int em_smart_pwr_down = 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
/* Controls whether promiscuous also shows bad packets */
static int em_debug_sbp = FALSE;
/* Local switch for MSI/MSIX */
static int em_enable_msi = TRUE;
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);
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);
- 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);
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);
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);
TUNABLE_INT("hw.em.enable_msi", &em_enable_msi);
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 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
#ifndef EM_LEGACY_IRQ
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
#endif
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* 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;
- 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 tsize, rsize;
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
u16 eeprom_data, device_id;
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));
EM_TX_LOCK_INIT(adapter, device_get_nameunit(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
EM_RX_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);
callout_init_mtx(&adapter->tx_fifo_timer, &adapter->tx_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) ||
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->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
(adapter->hw.mac.type == e1000_ich9lan)) {
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);
if (adapter->hw.mac.type >= e1000_82540) {
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_abs_int_delay",
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
"receive interrupt delay limit in usecs",
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->rx_abs_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_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
em_rx_abs_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_abs_int_delay",
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
"transmit interrupt delay limit in usecs",
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->tx_abs_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_TADV),
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
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
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
#ifndef EM_LEGACY_IRQ
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);
#endif
- 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 ||
(adapter->hw.mac.type >= e1000_82544 && em_txd > EM_MAX_TXD) ||
(adapter->hw.mac.type < e1000_82544 && em_txd > EM_MAX_TXD_82543) ||
- 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
(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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 ||
(adapter->hw.mac.type >= e1000_82544 && em_rxd > EM_MAX_RXD) ||
(adapter->hw.mac.type < e1000_82544 && em_rxd > EM_MAX_RXD_82543) ||
- 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
(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;
adapter->rx_buffer_len = 2048;
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
e1000_init_script_state_82541(&adapter->hw, TRUE);
e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
/* 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
tsize = roundup2(adapter->num_tx_desc * sizeof(struct e1000_tx_desc),
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);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Allocate Transmit Descriptor 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
if (em_dma_malloc(adapter, tsize, &adapter->txdma, BUS_DMA_NOWAIT)) {
- 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 tx_desc memory\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 = ENOMEM;
- 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
goto err_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
adapter->tx_desc_base =
(struct e1000_tx_desc *)adapter->txdma.dma_vaddr;
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
rsize = roundup2(adapter->num_rx_desc * sizeof(struct e1000_rx_desc),
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);
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
/* Allocate Receive Descriptor 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
if (em_dma_malloc(adapter, rsize, &adapter->rxdma, BUS_DMA_NOWAIT)) {
- 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_desc memory\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 = ENOMEM;
- 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
goto err_rx_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
adapter->rx_desc_base =
(struct e1000_rx_desc *)adapter->rxdma.dma_vaddr;
/* 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;
goto err_hw_init;
}
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Initialize the hardware */
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_hardware_init(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, "Unable to initialize the hardware\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;
- 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
goto err_hw_init;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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;
- 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
goto err_hw_init;
- 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;
goto err_hw_init;
}
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
/* Allocate transmit descriptors and buffers */
if (em_allocate_transmit_structures(adapter)) {
device_printf(dev, "Could not setup transmit structures\n");
error = ENOMEM;
goto err_tx_struct;
}
/* Allocate receive descriptors and buffers */
if (em_allocate_receive_structures(adapter)) {
device_printf(dev, "Could not setup receive structures\n");
error = ENOMEM;
goto err_rx_struct;
}
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
*/
if (adapter->msi > 1) /* Do MSI/X */
error = em_allocate_msix(adapter);
else /* MSI or Legacy */
error = em_allocate_legacy(adapter);
if (error)
goto err_rx_struct;
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
/* 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");
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Determine if we have to control management hardware */
adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
/*
* Setup Wake-on-Lan
*/
switch (adapter->hw.mac.type) {
case e1000_82542:
case e1000_82543:
break;
case e1000_82546:
case e1000_82546_rev_3:
case e1000_82571:
case e1000_80003es2lan:
if (adapter->hw.bus.func == 1)
e1000_read_nvm(&adapter->hw,
NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
else
e1000_read_nvm(&adapter->hw,
NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
eeprom_data &= EM_EEPROM_APME;
break;
default:
/* APME bit in EEPROM is mapped to WUC.APME */
eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC) &
E1000_WUC_APME;
break;
}
if (eeprom_data)
adapter->wol = E1000_WUFC_MAG;
/*
* 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_82546GB_PCIE:
adapter->wol = 0;
break;
case E1000_DEV_ID_82546EB_FIBER:
case E1000_DEV_ID_82546GB_FIBER:
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_82546GB_QUAD_COPPER_KSP3:
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;
}
/* Do we need workaround for 82544 PCI-X adapter? */
if (adapter->hw.bus.type == e1000_bus_type_pcix &&
adapter->hw.mac.type == e1000_82544)
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->pcix_82544 = TRUE;
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
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->pcix_82544 = FALSE;
- 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 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,
em_register_vlan, 0, EVENTHANDLER_PRI_FIRST);
adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
em_unregister_vlan, 0, EVENTHANDLER_PRI_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
/* Tell the stack that the interface is not active */
adapter->ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | 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
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);
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
err_rx_struct:
em_free_transmit_structures(adapter);
err_tx_struct:
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_hw_init:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
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(adapter, &adapter->rxdma);
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_rx_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
em_dma_free(adapter, &adapter->txdma);
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_tx_desc:
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_TX_LOCK_DESTROY(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_RX_LOCK_DESTROY(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 __FreeBSD_version >= 700000
if (adapter->ifp->if_vlantrunk != NULL) {
#else
if (adapter->ifp->if_nvlans != 0) {
#endif
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
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
EM_TX_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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
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
if (((adapter->hw.mac.type == e1000_82573) ||
(adapter->hw.mac.type == e1000_ich8lan) ||
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->hw.mac.type == e1000_ich10lan) ||
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
(adapter->hw.mac.type == e1000_ich9lan)) &&
e1000_check_mng_mode(&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
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->wol) {
E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
em_enable_wakeup(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
EM_TX_UNLOCK(adapter);
EM_CORE_UNLOCK(adapter);
Cleanup from __FreeBSD_version.
2005-10-14 10:34:46 +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);
callout_drain(&adapter->tx_fifo_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);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Free Transmit Descriptor 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
if (adapter->tx_desc_base) {
em_dma_free(adapter, &adapter->txdma);
adapter->tx_desc_base = 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
}
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
/* Free Receive Descriptor 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
if (adapter->rx_desc_base) {
em_dma_free(adapter, &adapter->rxdma);
adapter->rx_desc_base = 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
}
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_TX_LOCK_DESTROY(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_RX_LOCK_DESTROY(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);
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
EM_TX_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_stop(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
EM_TX_UNLOCK(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_manageability(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
if (((adapter->hw.mac.type == e1000_82573) ||
(adapter->hw.mac.type == e1000_ich8lan) ||
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->hw.mac.type == e1000_ich10lan) ||
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
(adapter->hw.mac.type == e1000_ich9lan)) &&
e1000_check_mng_mode(&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
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->wol) {
E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
em_enable_wakeup(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_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.
**********************************************************************/
static void
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
em_start_locked(struct ifnet *ifp)
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;
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;
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
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_TX_LOCK_ASSERT(adapter);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
Honor IFF_DRV_OACTIVE in em_start_locked().
2006-07-20 03:57:58 +00:00
if ((ifp->if_drv_flags & (IFF_DRV_RUNNING|IFF_DRV_OACTIVE)) !=
IFF_DRV_RUNNING)
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
if (!adapter->link_active)
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
return;
- 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
while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd)) {
Further refine the if_em vlan fix in if_em.c:1.53: - Because em_encap() can now fail in a way that leaves us without an mbuf chain, potentially set *m_headp to NULL if that happens, so that the caller can do the right thing. This case can occur when we try to prepend the vlan header mbuf but can't allocate additional memory. - Modify the caller of em_encap() to detect a NULL m_head and not try to queue the mbuf if that happens. - When em_encap() fails, make sure to call bus_dmamap_destroy() to clean up.
2004-11-14 20:20:28 +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
IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
if (m_head == NULL)
break;
Further refine the if_em vlan fix in if_em.c:1.53: - Because em_encap() can now fail in a way that leaves us without an mbuf chain, potentially set *m_headp to NULL if that happens, so that the caller can do the right thing. This case can occur when we try to prepend the vlan header mbuf but can't allocate additional memory. - Modify the caller of em_encap() to detect a NULL m_head and not try to queue the mbuf if that happens. - When em_encap() fails, make sure to call bus_dmamap_destroy() to clean up.
2004-11-14 20:20:28 +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
* Encapsulation can modify our pointer, and or make it
* NULL on failure. In that event, we can't requeue.
Further refine the if_em vlan fix in if_em.c:1.53: - Because em_encap() can now fail in a way that leaves us without an mbuf chain, potentially set *m_headp to NULL if that happens, so that the caller can do the right thing. This case can occur when we try to prepend the vlan header mbuf but can't allocate additional memory. - Modify the caller of em_encap() to detect a NULL m_head and not try to queue the mbuf if that happens. - When em_encap() fails, make sure to call bus_dmamap_destroy() to clean up.
2004-11-14 20:20:28 +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 (em_xmit(adapter, &m_head)) {
Further refine the if_em vlan fix in if_em.c:1.53: - Because em_encap() can now fail in a way that leaves us without an mbuf chain, potentially set *m_headp to NULL if that happens, so that the caller can do the right thing. This case can occur when we try to prepend the vlan header mbuf but can't allocate additional memory. - Modify the caller of em_encap() to detect a NULL m_head and not try to queue the mbuf if that happens. - When em_encap() fails, make sure to call bus_dmamap_destroy() to clean up.
2004-11-14 20:20:28 +00:00
if (m_head == NULL)
break;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
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_DRV_PREPEND(&ifp->if_snd, m_head);
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;
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
}
- 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
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
/* Send a copy of the frame to the BPF listener */
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
ETHER_BPF_MTAP(ifp, m_head);
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 timeout in case hardware has problems transmitting. */
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
adapter->watchdog_timer = EM_TX_TIMEOUT;
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
}
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)
{
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;
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_TX_LOCK(adapter);
Remove queue check from last commit. In most cases there is smth in queue, when start function is called. Reviewed by: ru
2005-09-20 14:52:57 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
Check IFF_DRV_RUNNING and presense of packets in queue before calling em_start_locked(). This fixes panic on shutdown with active traffic passing through router. Sponsored by: Rambler
2005-09-20 13:37:17 +00:00
em_start_locked(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
EM_TX_UNLOCK(adapter);
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
}
arp_ifinit(ifp, ifa);
} 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;
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 eeprom_data = 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
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_82573:
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
/*
* 82573 only supports jumbo frames
* if ASPM is disabled.
*/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_nvm(&adapter->hw,
NVM_INIT_3GIO_3, 1, &eeprom_data);
if (eeprom_data & NVM_WORD1A_ASPM_MASK) {
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 = ETHER_MAX_LEN;
break;
}
/* Allow Jumbo frames - fall thru */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_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:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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: /* Limit 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;
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 */
case e1000_82542:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
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
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
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
EM_TX_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_stop(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
EM_TX_UNLOCK(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->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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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) {
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);
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
}
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
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;
}
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
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
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;
}
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);
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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
VLAN_CAPABILITIES(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
#endif
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
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
#ifdef EM_TIMESYNC
/*
** IOCTL support for Precision Time (IEEE 1588) Support
*/
case EM_TIMESYNC_READTS:
{
u32 rx_ctl, tx_ctl;
struct em_tsync_read *tdata;
tdata = (struct em_tsync_read *) ifr->ifr_data;
IOCTL_DEBUGOUT("Reading Timestamp\n");
if (tdata->read_current_time) {
getnanotime(&tdata->system_time);
tdata->network_time = E1000_READ_REG(&adapter->hw, E1000_SYSTIML);
tdata->network_time |=
(u64)E1000_READ_REG(&adapter->hw, E1000_SYSTIMH ) << 32;
}
rx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCRXCTL);
tx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCTXCTL);
IOCTL_DEBUGOUT1("RX_CTL value = %u\n", rx_ctl);
IOCTL_DEBUGOUT1("TX_CTL value = %u\n", tx_ctl);
if (rx_ctl & 0x1) {
IOCTL_DEBUGOUT("RX timestamp is valid\n");
u32 tmp;
unsigned char *tmp_cp;
tdata->rx_valid = 1;
tdata->rx_stamp = E1000_READ_REG(&adapter->hw, E1000_RXSTMPL);
tdata->rx_stamp |= (u64)E1000_READ_REG(&adapter->hw,
E1000_RXSTMPH) << 32;
tmp = E1000_READ_REG(&adapter->hw, E1000_RXSATRL);
tmp_cp = (unsigned char *) &tmp;
tdata->srcid[0] = tmp_cp[0];
tdata->srcid[1] = tmp_cp[1];
tdata->srcid[2] = tmp_cp[2];
tdata->srcid[3] = tmp_cp[3];
tmp = E1000_READ_REG(&adapter->hw, E1000_RXSATRH);
tmp_cp = (unsigned char *) &tmp;
tdata->srcid[4] = tmp_cp[0];
tdata->srcid[5] = tmp_cp[1];
tdata->seqid = tmp >> 16;
tdata->seqid = htons(tdata->seqid);
} else
tdata->rx_valid = 0;
if (tx_ctl & 0x1) {
IOCTL_DEBUGOUT("TX timestamp is valid\n");
tdata->tx_valid = 1;
tdata->tx_stamp = E1000_READ_REG(&adapter->hw, E1000_TXSTMPL);
tdata->tx_stamp |= (u64) E1000_READ_REG(&adapter->hw,
E1000_TXSTMPH) << 32;
} else
tdata->tx_valid = 0;
return (0);
}
#endif /* EM_TIMESYNC */
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
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
}
/*********************************************************************
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
* Watchdog timer:
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
*
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
* This routine is called from the local timer every second.
* As long as transmit descriptors are being cleaned the value
* is non-zero and we do nothing. Reaching 0 indicates a tx hang
* and we then reset the device.
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
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
em_watchdog(struct adapter *adapter)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
EM_CORE_LOCK_ASSERT(adapter);
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
/*
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
** The timer is set to 5 every time start queues a packet.
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
** Then txeof keeps resetting it as long as it cleans at
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
** least one descriptor.
** Finally, anytime all descriptors are clean the timer is
** set to 0.
*/
This update primarily addresses the ability to have both the em and the igb driver static in the kernel. But it also reflects some other bug fixes in my development stream at Intel. PR 122373 is also fixed in this code.
2008-04-02 22:00:36 +00:00
EM_TX_LOCK(adapter);
if ((adapter->watchdog_timer == 0) || (--adapter->watchdog_timer)) {
EM_TX_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
return;
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
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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
/* If we are in this routine because of pause frames, then
* don't reset the hardware.
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
if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
E1000_STATUS_TXOFF) {
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
adapter->watchdog_timer = EM_TX_TIMEOUT;
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
EM_TX_UNLOCK(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
if (e1000_check_for_link(&adapter->hw) == 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
device_printf(adapter->dev, "watchdog timeout -- resetting\n");
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
adapter->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
adapter->watchdog_events++;
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
EM_TX_UNLOCK(adapter);
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
em_init_locked(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
}
/*********************************************************************
* 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
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
EM_TX_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_stop(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
EM_TX_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
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 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
*
* Devices before the 82547 had a Packet Buffer of 64K.
* Default allocation: PBA=48K for Rx, leaving 16K for Tx.
* After the 82547 the buffer was reduced to 40K.
* Default allocation: PBA=30K for Rx, leaving 10K for Tx.
* Note: default does not leave enough room for Jumbo Frame >10k.
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 the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_82547:
case e1000_82547_rev_2: /* 82547: Total Packet Buffer is 40K */
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)
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
pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
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
else
pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
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->tx_fifo_head = 0;
adapter->tx_head_addr = pba << EM_TX_HEAD_ADDR_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
adapter->tx_fifo_size =
(E1000_PBA_40K - pba) << EM_PBA_BYTES_SHIFT;
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* 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:
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:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_PBA_10K 0x000A
pba = E1000_PBA_10K;
break;
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:
/* Devices before 82547 had a Packet Buffer of 64K. */
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 */
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_hardware_init(adapter)) {
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
device_printf(dev, "Unable to initialize the hardware\n");
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_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);
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
/* New register interface replaces this but
waiting on kernel support to be added */
if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) &&
((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0)) {
u32 ctrl;
ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
ctrl |= E1000_CTRL_VME;
E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
}
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
/* 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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if (adapter->hw.mac.type >= e1000_82543) {
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
if (ifp->if_capenable & IFCAP_TXCSUM)
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
ifp->if_hwassist |= (CSUM_TCP | CSUM_UDP);
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
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
if (ifp->if_capenable & IFCAP_TSO4)
ifp->if_hwassist |= CSUM_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
#endif
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");
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
EM_TX_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_stop(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
EM_TX_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
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);
/*
** Set the IVAR - interrupt vector routing.
** Each nibble represents a vector, high bit
** is enable, other 3 bits are the MSIX table
** entry, we map RXQ0 to 0, TXQ0 to 1, and
** Link (other) to 2, hence the magic number.
*/
E1000_WRITE_REG(&adapter->hw, E1000_IVAR, 0x800A0908);
}
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
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
#ifdef EM_TIMESYNC
/* Initializae IEEE 1588 Precision Time hardware */
if ((adapter->hw.mac.type == e1000_82574) ||
(adapter->hw.mac.type == e1000_ich10lan))
em_tsync_init(adapter);
#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
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
/*********************************************************************
*
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
* Legacy polling routine
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
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
{
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;
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;
- 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);
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
return;
}
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)) {
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_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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +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);
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_rxeof(adapter, count);
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_TX_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_txeof(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
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
em_start_locked(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
EM_TX_UNLOCK(adapter);
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 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
#ifdef EM_LEGACY_IRQ
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
/*********************************************************************
*
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
* Legacy Interrupt Service routine
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
*
*********************************************************************/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 void
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_intr(void *arg)
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +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
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
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_icr;
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
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 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_capenable & IFCAP_POLLING)
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
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
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
for (;;) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
if (adapter->hw.mac.type >= e1000_82571 &&
(reg_icr & E1000_ICR_INT_ASSERTED) == 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
break;
else if (reg_icr == 0)
break;
/*
* XXX: some laptops trigger several spurious interrupts
* on em(4) when in the resume cycle. The ICR register
* reports all-ones value in this case. Processing such
* interrupts would lead to a freeze. I don't know why.
*/
if (reg_icr == 0xffffffff)
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
EM_CORE_UNLOCK(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) {
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_rxeof(adapter, -1);
Driver version 6.7.3 - Bring HEAD up to the latest shared code - Fix TSO problem using limited MSS and forwarding - Dual lock implementation - New device support - For my ease, this code can compile in either 6.x or later - brings this driver in sync with the 6.3
2007-11-20 21:41:22 +00:00
EM_TX_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_txeof(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_TX_UNLOCK(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
}
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_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
/* Link status change */
if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
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_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);
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
/* Deal with TX cruft when link lost */
em_tx_purge(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
callout_reset(&adapter->timer, hz,
em_local_timer, 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 (reg_icr & E1000_ICR_RXO)
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->rx_overruns++;
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
}
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);
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 &&
!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
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_start(ifp);
Add polling support to the "em" driver. Reviewed by: pdeuskar (maintainer) Approved by: re
2002-11-25 19:56:16 +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
#else /* EM_FAST_IRQ, then fast interrupt routines only */
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)
{
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;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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 (!(ifp->if_drv_flags & IFF_DRV_RUNNING))
- 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
return;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
EM_CORE_LOCK(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
callout_stop(&adapter->timer);
em_update_link_status(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
/* Deal with TX cruft when link lost */
em_tx_purge(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
callout_reset(&adapter->timer, hz, em_local_timer, 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);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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 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
/* Combined RX/TX handler, used by Legacy and MSI */
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
static void
em_handle_rxtx(void *context, int pending)
{
struct adapter *adapter = context;
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 ifnet *ifp = adapter->ifp;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
if (em_rxeof(adapter, adapter->rx_process_limit) != 0)
taskqueue_enqueue(adapter->tq, &adapter->rxtx_task);
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_TX_LOCK(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_txeof(adapter);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
em_start_locked(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
EM_TX_UNLOCK(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
}
em_enable_intr(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 void
em_handle_rx(void *context, int pending)
{
struct adapter *adapter = context;
struct ifnet *ifp = adapter->ifp;
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) &&
(em_rxeof(adapter, adapter->rx_process_limit) != 0))
taskqueue_enqueue(adapter->tq, &adapter->rx_task);
}
static void
em_handle_tx(void *context, int pending)
{
struct adapter *adapter = context;
struct ifnet *ifp = adapter->ifp;
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
EM_TX_LOCK(adapter);
em_txeof(adapter);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
em_start_locked(ifp);
EM_TX_UNLOCK(adapter);
}
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
*
*********************************************************************/
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
#define FILTER_STRAY
#define FILTER_HANDLED
static void
#else
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
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
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
/*
* Mask interrupts until the taskqueue is finished running. This is
* cheap, just assume that it is needed. This also works around the
* MSI message reordering errata on certain systems.
*/
em_disable_intr(adapter);
taskqueue_enqueue(adapter->tq, &adapter->rxtx_task);
/* 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
/*********************************************************************
*
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 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
#define EM_MSIX_TX 0x00040000
#define EM_MSIX_RX 0x00010000
#define EM_MSIX_LINK 0x00100000
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
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
{
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
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
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->tx_irq;
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
EM_TX_LOCK(adapter);
em_txeof(adapter);
EM_TX_UNLOCK(adapter);
taskqueue_enqueue(adapter->tq, &adapter->tx_task);
}
/* Reenable this interrupt */
E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_TX);
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 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 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
*
**********************************************************************/
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_rx(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
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
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->rx_irq;
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) &&
(em_rxeof(adapter, adapter->rx_process_limit) != 0))
taskqueue_enqueue(adapter->tq, &adapter->rx_task);
/* Reenable this interrupt */
E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_RX);
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;
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
taskqueue_enqueue(taskqueue_fast, &adapter->link_task);
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(&adapter->hw, E1000_IMS,
EM_MSIX_LINK | E1000_IMS_LSC);
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
}
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 /* EM_FAST_IRQ */
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
if (adapter->hw.mac.type == e1000_82545)
fiber_type = IFM_1000_LX;
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
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_xmit(struct adapter *adapter, 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
{
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;
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
struct m_tag *mtag;
#endif
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;
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
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);
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
do_tso = 0;
#endif
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/*
* Force a cleanup if number of TX descriptors
* available hits the threshold
*/
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->num_tx_desc_avail <= EM_TX_CLEANUP_THRESHOLD) {
em_txeof(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
/* Now do we at least have a minimal? */
if (adapter->num_tx_desc_avail <= EM_TX_OP_THRESHOLD) {
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->no_tx_desc_avail1++;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
return (ENOBUFS);
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
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
}
This is the merge of the Intel 6.2.9 driver. It provides all new shared code, new device support, and it is hoped a more stable driver for 6.2. RELEASE. This checkin was discussed and approved today by RE, scottl, jhb, and pdeuskar
2006-10-28 00:47:55 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +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
*/
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 = adapter->next_avail_tx_desc;
tx_buffer = &adapter->tx_buffer_area[first];
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 = bus_dmamap_load_mbuf_sg(adapter->txtag, map,
*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 */
error = bus_dmamap_load_mbuf_sg(adapter->txtag, map,
*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
if (error == ENOMEM) {
adapter->no_tx_dma_setup++;
return (error);
} else if (error != 0) {
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 == ENOMEM) {
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->no_tx_dma_setup++;
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
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++;
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
m_freem(*m_headp);
*m_headp = NULL;
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
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.
*/
if ((do_tso == 0) && (adapter->tx_tso == TRUE)) {
if (nsegs == 1)
tso_desc = TRUE;
adapter->tx_tso = FALSE;
}
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
if (nsegs > (adapter->num_tx_desc_avail - 2)) {
adapter->no_tx_desc_avail2++;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
bus_dmamap_unload(adapter->txtag, map);
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) {
error = em_tso_setup(adapter, m_head, &txd_upper, &txd_lower);
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 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
#ifndef EM_TIMESYNC
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
/*
** Timesync needs to check the packet header
** so call checksum code to do so, but don't
** penalize the code if not defined.
*/
if (m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD)
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
#endif
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
em_transmit_checksum_setup(adapter, m_head,
&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
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
i = adapter->next_avail_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
if (adapter->pcix_82544)
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
txd_saved = i;
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 is 82544 and on PCIX bus */
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->pcix_82544) {
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
DESC_ARRAY desc_array;
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 array_elements, counter;
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
/*
* Check the Address and Length combination and
* split the data accordingly
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
*/
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
array_elements = em_fill_descriptors(segs[j].ds_addr,
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
segs[j].ds_len, &desc_array);
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
for (counter = 0; counter < array_elements; counter++) {
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 (txd_used == adapter->num_tx_desc_avail) {
adapter->next_avail_tx_desc = txd_saved;
adapter->no_tx_desc_avail2++;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
bus_dmamap_unload(adapter->txtag, map);
return (ENOBUFS);
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
}
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
tx_buffer = &adapter->tx_buffer_area[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
ctxd = &adapter->tx_desc_base[i];
ctxd->buffer_addr = htole64(
desc_array.descriptor[counter].address);
ctxd->lower.data = htole32(
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 | txd_lower | (u16)
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
desc_array.descriptor[counter].length));
ctxd->upper.data =
htole32((txd_upper));
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 = i;
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_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
i = 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
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;
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
txd_used++;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
}
Add 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 {
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
tx_buffer = &adapter->tx_buffer_area[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
ctxd = &adapter->tx_desc_base[i];
Fix TSO support on sun4v - incorporate csjp's fix for a mishandled endian conversion - convert PAGE_SIZE to 4096 for PCIe adapter workaround (my page size is not 4k) - implement em_read_pcie_cap_reg where we set the max read size on pcie to 4k (taken from mxge) Reviewed by: scottl and jfvogel
2006-11-23 00:52:52 +00:00
seg_addr = segs[j].ds_addr;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->buffer_addr = htole64(seg_addr);
ctxd->lower.data = htole32(
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
adapter->txd_cmd | txd_lower | seg_len);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.data =
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
htole32(txd_upper);
if (++i == adapter->num_tx_desc)
i = 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
/* Now make the sentinel */
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
++txd_used; /* using an extra txd */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 = &adapter->tx_desc_base[i];
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
tx_buffer = &adapter->tx_buffer_area[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
ctxd->buffer_addr =
htole64(seg_addr + seg_len);
ctxd->lower.data = htole32(
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
adapter->txd_cmd | txd_lower | 4);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.data =
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
htole32(txd_upper);
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 = i;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
if (++i == adapter->num_tx_desc)
i = 0;
} else {
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
ctxd->buffer_addr = htole64(seg_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
ctxd->lower.data = htole32(
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
adapter->txd_cmd | txd_lower | seg_len);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.data =
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
htole32(txd_upper);
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 = i;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
if (++i == adapter->num_tx_desc)
i = 0;
}
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;
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
}
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
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->next_avail_tx_desc = i;
if (adapter->pcix_82544)
adapter->num_tx_desc_avail -= txd_used;
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
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_avail -= nsegs;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
if (tso_desc) /* TSO used an extra for sentinel */
adapter->num_tx_desc_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
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
/*
** Handle VLAN tag, this is the
** biggest difference between
** 6.x and 7
*/
#if __FreeBSD_version < 700000
/* Find out if we are in vlan mode. */
mtag = VLAN_OUTPUT_TAG(ifp, m_head);
if (mtag != NULL) {
ctxd->upper.fields.special =
htole16(VLAN_TAG_VALUE(mtag));
#else /* FreeBSD 7 */
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);
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_dmamap_sync(adapter->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
*/
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 = &adapter->tx_buffer_area[first];
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.
*/
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
bus_dmamap_sync(adapter->txdma.dma_tag, adapter->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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_82547 &&
Small style and comment adjustments. Reviewed by: jfv
2006-09-29 13:54:04 +00:00
adapter->link_duplex == HALF_DUPLEX)
- 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
em_82547_move_tail(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 {
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_TDT(0), 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
if (adapter->hw.mac.type == e1000_82547)
em_82547_update_fifo_head(adapter,
m_head->m_pkthdr.len);
}
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
#ifdef EM_TIMESYNC
if (ctxd->upper.data & E1000_TXD_EXTCMD_TSTAMP) {
HW_DEBUGOUT( "@@@ Timestamp bit is set in transmit descriptor\n" );
}
#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
return (0);
}
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
/*********************************************************************
*
* 82547 workaround to avoid controller hang in half-duplex environment.
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
* The workaround is to avoid queuing a large packet that would span
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
* the internal Tx FIFO ring boundary. We need to reset the FIFO pointers
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
* in this case. We do that only when FIFO is quiescent.
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
*
**********************************************************************/
static void
- 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
em_82547_move_tail(void *arg)
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
{
- 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
struct adapter *adapter = arg;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 *tx_desc;
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 hw_tdt, sw_tdt, length = 0;
bool eop = 0;
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
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_TX_LOCK_ASSERT(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
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
hw_tdt = E1000_READ_REG(&adapter->hw, E1000_TDT(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
sw_tdt = adapter->next_avail_tx_desc;
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
while (hw_tdt != sw_tdt) {
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
tx_desc = &adapter->tx_desc_base[hw_tdt];
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
length += tx_desc->lower.flags.length;
eop = tx_desc->lower.data & E1000_TXD_CMD_EOP;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 (++hw_tdt == adapter->num_tx_desc)
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
hw_tdt = 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
if (eop) {
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_82547_fifo_workaround(adapter, length)) {
adapter->tx_fifo_wrk_cnt++;
callout_reset(&adapter->tx_fifo_timer, 1,
em_82547_move_tail, adapter);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
break;
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
}
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_TDT(0), hw_tdt);
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_82547_update_fifo_head(adapter, length);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +00:00
length = 0;
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
}
}
}
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_82547_fifo_workaround(struct adapter *adapter, int len)
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
{
int fifo_space, fifo_pkt_len;
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
fifo_pkt_len = roundup2(len + EM_FIFO_HDR, EM_FIFO_HDR);
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
if (adapter->link_duplex == HALF_DUPLEX) {
fifo_space = adapter->tx_fifo_size - adapter->tx_fifo_head;
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 (fifo_pkt_len >= (EM_82547_PKT_THRESH + fifo_space)) {
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_82547_tx_fifo_reset(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);
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
- 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 (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
}
}
- 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);
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
}
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_82547_update_fifo_head(struct adapter *adapter, int len)
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
{
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
int fifo_pkt_len = roundup2(len + EM_FIFO_HDR, EM_FIFO_HDR);
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
/* tx_fifo_head is always 16 byte aligned */
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->tx_fifo_head += fifo_pkt_len;
if (adapter->tx_fifo_head >= adapter->tx_fifo_size) {
adapter->tx_fifo_head -= adapter->tx_fifo_size;
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
}
}
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_82547_tx_fifo_reset(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
{
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;
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
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 ((E1000_READ_REG(&adapter->hw, E1000_TDT(0)) ==
E1000_READ_REG(&adapter->hw, E1000_TDH(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_READ_REG(&adapter->hw, E1000_TDFT) ==
E1000_READ_REG(&adapter->hw, E1000_TDFH)) &&
(E1000_READ_REG(&adapter->hw, E1000_TDFTS) ==
E1000_READ_REG(&adapter->hw, E1000_TDFHS)) &&
(E1000_READ_REG(&adapter->hw, E1000_TDFPC) == 0)) {
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
/* Disable TX 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
tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
E1000_WRITE_REG(&adapter->hw, E1000_TCTL,
tctl & ~E1000_TCTL_EN);
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
/* Reset FIFO pointers */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_TDFT,
adapter->tx_head_addr);
E1000_WRITE_REG(&adapter->hw, E1000_TDFH,
adapter->tx_head_addr);
E1000_WRITE_REG(&adapter->hw, E1000_TDFTS,
adapter->tx_head_addr);
E1000_WRITE_REG(&adapter->hw, E1000_TDFHS,
adapter->tx_head_addr);
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
/* Re-enable TX 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);
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_FLUSH(&adapter->hw);
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
adapter->tx_fifo_head = 0;
adapter->tx_fifo_reset_cnt++;
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
- 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);
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
}
else {
- 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 (FALSE);
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
}
}
- 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;
u8 mta[512]; /* Largest MTS is 4096 bits */
- 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
IF_ADDR_LOCK(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
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++;
}
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
IF_ADDR_UNLOCK(ifp);
- 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
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
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_update_mc_addr_list(&adapter->hw, mta,
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
mcnt, 1, adapter->hw.mac.rar_entry_count);
- 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
}
- 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;
- 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
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_smartspeed(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
/*
* Each second we check the watchdog to
* protect against hardware hangs.
*/
em_watchdog(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
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);
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 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 */
adapter->watchdog_timer = FALSE;
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;
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);
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
EM_TX_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);
callout_stop(&adapter->tx_fifo_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
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
#ifdef EM_TIMESYNC
/* Disable IEEE 1588 Time hardware */
if ((adapter->hw.mac.type == e1000_82574) ||
(adapter->hw.mac.type == e1000_ich10lan))
em_tsync_disable(adapter);
#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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
if (adapter->hw.mac.type >= e1000_82544)
E1000_WRITE_REG(&adapter->hw, E1000_WUC, 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
/*********************************************************************
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;
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, rid, error = E1000_SUCCESS;
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Only older adapters use IO mapping */
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.mac.type > e1000_82543) &&
Mistake in the logic deciding what adapters need to map the IO BAR. Causing the driver to fail on th 82542. Reviewed by:pdeuskar Approved by:pdeuskar
2007-05-11 21:36:08 +00:00
(adapter->hw.mac.type < e1000_82571)) {
- 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
/* Figure our where our IO BAR is ? */
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
for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
- 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
val = pci_read_config(dev, rid, 4);
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
if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
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->io_rid = rid;
- 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
break;
}
rid += 4;
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
/* check for 64bit BAR */
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
if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
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 += 4;
}
if (rid >= PCIR_CIS) {
- 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 locate IO BAR\n");
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
return (ENXIO);
- 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
adapter->ioport = 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_IOPORT, &adapter->io_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->ioport == 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: "
"ioport\n");
return (ENXIO);
- 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
}
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.io_base = 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
adapter->osdep.io_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->ioport);
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.io_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->ioport);
}
/*
** Init the resource arrays
** used by MSIX setup
*/
for (int i = 0; i < 3; i++) {
adapter->rid[i] = i + 1; /* MSI/X RID starts at 1 */
adapter->tag[i] = NULL;
adapter->res[i] = NULL;
- Use IO mode to reset the controller (82544 and beyond) - Read the Mac address only once during attach. (This fixes the failover issue observed using the bonding driver) MFC after: 3 days
2002-07-16 16:55:03 +00:00
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
*/
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 (em_enable_msi)
adapter->msi = 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
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 (error);
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
*
**********************************************************************/
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;
int error;
/* 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
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
/* Legacy RID is 0 */
if (adapter->msi == 0)
adapter->rid[0] = 0;
/* We allocate a single interrupt resource */
adapter->res[0] = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &adapter->rid[0], RF_SHAREABLE | RF_ACTIVE);
if (adapter->res[0] == NULL) {
device_printf(dev, "Unable to allocate bus resource: "
"interrupt\n");
return (ENXIO);
}
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
#ifdef EM_LEGACY_IRQ
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* We do Legacy setup */
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 ((error = bus_setup_intr(dev, adapter->res[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
#if __FreeBSD_version > 700000
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
INTR_TYPE_NET | INTR_MPSAFE, NULL, em_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
#else /* 6.X */
INTR_TYPE_NET | INTR_MPSAFE, em_intr, adapter,
#endif
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->tag[0])) != 0) {
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
device_printf(dev, "Failed to register interrupt handler");
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
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 /* FAST_IRQ */
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/*
* Try allocating a fast interrupt and the associated deferred
* processing contexts.
*/
TASK_INIT(&adapter->rxtx_task, 0, em_handle_rxtx, adapter);
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));
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
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 ((error = bus_setup_intr(dev, adapter->res[0],
INTR_TYPE_NET | INTR_FAST, em_irq_fast, 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
#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
if ((error = bus_setup_intr(dev, adapter->res[0],
INTR_TYPE_NET, em_irq_fast, NULL, 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
#endif
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->tag[0])) != 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 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
#endif /* EM_LEGACY_IRQ */
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
{
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
int error;
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);
/* First get the resources */
for (int i = 0; i < adapter->msi; i++) {
adapter->res[i] = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &adapter->rid[i], RF_ACTIVE);
if (adapter->res[i] == NULL) {
device_printf(dev,
"Unable to allocate bus resource: "
"MSIX Interrupt\n");
return (ENXIO);
}
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
/*
* Now allocate deferred processing contexts.
*/
TASK_INIT(&adapter->rx_task, 0, em_handle_rx, adapter);
TASK_INIT(&adapter->tx_task, 0, em_handle_tx, adapter);
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));
/*
* And setup the interrupt handlers
*/
/* First slot to RX */
if ((error = bus_setup_intr(dev, adapter->res[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
#if __FreeBSD_version > 700000
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
INTR_TYPE_NET | INTR_MPSAFE, NULL, em_msix_rx, adapter,
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
#else /* 6.X */
INTR_TYPE_NET | INTR_MPSAFE, em_msix_rx, adapter,
#endif
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->tag[0])) != 0) {
device_printf(dev, "Failed to register RX handler");
return (error);
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
/* Next TX */
if ((error = bus_setup_intr(dev, adapter->res[1],
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
#if __FreeBSD_version > 700000
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
INTR_TYPE_NET | INTR_MPSAFE, NULL, em_msix_tx, adapter,
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
#else /* 6.X */
INTR_TYPE_NET | INTR_MPSAFE, em_msix_tx, adapter,
#endif
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->tag[1])) != 0) {
device_printf(dev, "Failed to register TX handler");
return (error);
}
/* And Link */
if ((error = bus_setup_intr(dev, adapter->res[2],
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
#if __FreeBSD_version > 700000
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
INTR_TYPE_NET | INTR_MPSAFE, NULL, em_msix_link, adapter,
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
#else /* 6.X */
INTR_TYPE_NET | INTR_MPSAFE, em_msix_link, adapter,
#endif
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->tag[2])) != 0) {
device_printf(dev, "Failed to register TX handler");
return (error);
}
return (0);
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)
{
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
/* Make sure the for loop below runs once */
if (adapter->msi == 0)
adapter->msi = 1;
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
/*
* First release all the interrupt resources:
* notice that since these are just kept
* in an array we can do the same logic
* whether its MSIX or just legacy.
*/
for (int i = 0; i < adapter->msi; i++) {
if (adapter->tag[i] != NULL) {
bus_teardown_intr(dev, adapter->res[i],
adapter->tag[i]);
adapter->tag[i] = NULL;
}
if (adapter->res[i] != NULL) {
bus_release_resource(dev, SYS_RES_IRQ,
adapter->rid[i], adapter->res[i]);
}
}
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
if (adapter->msi)
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
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->msix != NULL)
bus_release_resource(dev, SYS_RES_MEMORY,
PCIR_BAR(EM_MSIX_BAR), adapter->msix);
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);
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 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->ioport != 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_IOPORT,
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
adapter->io_rid, adapter->ioport);
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
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 MSI/X
*/
This change introduces a split to the Intel E1000 driver, now rather than just em, there is an igb driver (this follows behavior with our Linux drivers). All adapters up to the 82575 are supported in em, and new client/desktop support will continue to be in that adapter. The igb driver is for new server NICs like the 82575 and its followons. Advanced features for virtualization and performance will be in this driver. Also, both drivers now have shared code that is up to the latest we have released. Some stylistic changes as well. Enjoy :)
2008-02-29 21:50:11 +00:00
static int
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
if (adapter->hw.mac.type < e1000_82571)
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
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 */
if (adapter->hw.mac.type == e1000_82574) {
/* Map the MSIX BAR */
int rid = PCIR_BAR(EM_MSIX_BAR);
adapter->msix = bus_alloc_resource_any(dev,
SYS_RES_MEMORY, &rid, RF_ACTIVE);
if (!adapter->msix) {
/* May not be enabled */
device_printf(adapter->dev,
"Unable to map MSIX table \n");
goto msi;
}
val = pci_msix_count(dev);
/*
** 82574 can be configured for 5 but
** we limit use to 3.
*/
if (val > 3) val = 3;
if ((val) && pci_alloc_msix(dev, &val) == 0) {
device_printf(adapter->dev,"Using MSIX interrupts\n");
return (val);
}
}
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) {
adapter->msi = 1;
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
}
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
}
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
*
**********************************************************************/
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_hardware_init(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;
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 rx_buffer_size;
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
- 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_hardware_init: 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
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Issue a global reset */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
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 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
/* Get control from any management/hw control */
if (((adapter->hw.mac.type == e1000_82573) ||
(adapter->hw.mac.type == e1000_ich8lan) ||
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->hw.mac.type == e1000_ich10lan) ||
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
(adapter->hw.mac.type == e1000_ich9lan)) &&
e1000_check_mng_mode(&adapter->hw))
em_get_hw_control(adapter);
Added support for 82541 and 82547 based adapters. - These have Intel gigabit PHY - 82547 uses CSA interface MFC after: 1 week
2003-03-21 21:47:31 +00:00
/* When hardware is reset, fifo_head is also reset */
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->tx_fifo_head = 0;
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. */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_smart_pwr_down && (adapter->hw.mac.type == e1000_82571 ||
adapter->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. */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_phy_reg(&adapter->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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_phy_reg(&adapter->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.
*/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
rx_buffer_size = ((E1000_READ_REG(&adapter->hw, E1000_PBA) &
0xffff) << 10 );
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 = rx_buffer_size -
roundup2(adapter->max_frame_size, 1024);
adapter->hw.fc.low_water = adapter->hw.fc.high_water - 1500;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_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
adapter->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
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.pause_time = EM_FC_PAUSE_TIME;
adapter->hw.fc.send_xon = TRUE;
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->hw.fc.requested_mode = e1000_fc_full;
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_init_hw(&adapter->hw) < 0) {
device_printf(dev, "Hardware Initialization Failed\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
return (EIO);
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
e1000_check_for_link(&adapter->hw);
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 (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
}
/*********************************************************************
*
* 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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if (adapter->hw.mac.type >= e1000_82543) {
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
int version_cap;
#if __FreeBSD_version < 700000
version_cap = IFCAP_HWCSUM;
#else
version_cap = IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM;
#endif
ifp->if_capabilities |= version_cap;
ifp->if_capenable |= version_cap;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
- 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
#if __FreeBSD_version >= 700000
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
/* Identify TSO capable adapters */
if ((adapter->hw.mac.type > e1000_82544) &&
(adapter->hw.mac.type != e1000_82547))
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
ifp->if_capabilities |= IFCAP_TSO4;
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
/*
* By default only enable on PCI-E, this
* can be overriden by ifconfig.
*/
if (adapter->hw.mac.type >= e1000_82571)
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
ifp->if_capenable |= IFCAP_TSO4;
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
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
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
/*
* Tell the upper layer(s) we support long frames.
*/
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;
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
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
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
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
if (adapter->hw.mac.type == e1000_82545)
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
fiber_type = IFM_1000_LX;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
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
}
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
/*********************************************************************
*
* Workaround for SmartSpeed on 82541 and 82547 controllers
*
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 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
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_smartspeed(struct adapter *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
{
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;
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
if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
adapter->hw.mac.autoneg == 0 ||
(adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
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
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
if (adapter->smartspeed == 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
/* If Master/Slave config fault is asserted twice,
* we assume back-to-back */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
- 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 (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
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
e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
- 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 (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
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
e1000_read_phy_reg(&adapter->hw,
PHY_1000T_CTRL, &phy_tmp);
- 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(phy_tmp & CR_1000T_MS_ENABLE) {
phy_tmp &= ~CR_1000T_MS_ENABLE;
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
e1000_write_phy_reg(&adapter->hw,
PHY_1000T_CTRL, phy_tmp);
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->smartspeed++;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.autoneg &&
!e1000_phy_setup_autoneg(&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
!e1000_read_phy_reg(&adapter->hw,
PHY_CONTROL, &phy_tmp)) {
- 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
phy_tmp |= (MII_CR_AUTO_NEG_EN |
MII_CR_RESTART_AUTO_NEG);
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
e1000_write_phy_reg(&adapter->hw,
PHY_CONTROL, phy_tmp);
- 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;
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
} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
- 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 still no link, perhaps using 2/3 pair cable */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
- 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
phy_tmp |= CR_1000T_MS_ENABLE;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
if(adapter->hw.mac.autoneg &&
!e1000_phy_setup_autoneg(&adapter->hw) &&
!e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
- 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
phy_tmp |= (MII_CR_AUTO_NEG_EN |
MII_CR_RESTART_AUTO_NEG);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
}
}
/* Restart process after EM_SMARTSPEED_MAX iterations */
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->smartspeed++ == EM_SMARTSPEED_MAX)
adapter->smartspeed = 0;
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
}
- 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;
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
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 */
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
error = bus_dma_tag_create(NULL, /* parent */
#endif
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
}
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
* the information needed to transmit a packet on the wire.
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_transmit_structures(struct adapter *adapter)
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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;
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
struct em_buffer *tx_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
int error;
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
/*
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 DMA tags for tx descriptors
- 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
#if __FreeBSD_version >= 700000
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 ((error = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
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
if ((error = bus_dma_tag_create(NULL, /* parent */
#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
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
EM_TSO_SIZE, /* 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
EM_MAX_SCATTER, /* 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
EM_TSO_SEG_SIZE, /* maxsegsize */
- 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
0, /* flags */
NULL, /* lockfunc */
NULL, /* lockarg */
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->txtag)) != 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 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
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->tx_buffer_area = malloc(sizeof(struct em_buffer) *
adapter->num_tx_desc, M_DEVBUF, M_NOWAIT | M_ZERO);
if (adapter->tx_buffer_area == NULL) {
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
/* Create the descriptor buffer dma maps */
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
for (int i = 0; i < adapter->num_tx_desc; i++) {
tx_buffer = &adapter->tx_buffer_area[i];
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 = bus_dmamap_create(adapter->txtag, 0, &tx_buffer->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;
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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;
busdma cleanup for em(4). - don't force busdma to pre-allocate bounce pages for parent tag. - use system supplied roundup2 macro instead of rolling its own version. - TX/RX decriptor length should be multiple of 128. There is no no need to expand the size with the multiple of 4096. - don't create/destroy DMA maps in TX/RX handlers. Use pre-allocated DMA maps. Since creating DMA maps on sparc64 is time consuming operations(resource mananger overhead), this change should boost performance on sparc64. I could get > 2x speedup on Ultra60. - TX/RX descriptors could be aligned on 128 boundary. Aligning them on PAGE_SIZE is waste of resource. - don't blindly create TX DMA tag with size of MCLBYTES * 8. The size is only valid under jumbo frame environments. Instead of using the hardcoded value, re-compute necessary size on the fly. - RX side bus_dmamap_load_mbuf_sg(9) support. - remove unused macro EM_ROUNDUP and constant EM_MMBA. Reviewed by: scottl Tested by: glebius
2005-11-21 04:17:43 +00:00
}
- Bus DMA'fy the driver - Use htole* macros where appropriate so that the driver could work on non-x86 architectures - Use m_getcl() instead of MGETHDR/MCLGET macros Submitted by: sam (Sam Leffler)
2003-05-02 21:17:08 +00:00
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 (0);
fail:
em_free_transmit_structures(adapter);
return (error);
}
/*********************************************************************
*
* (Re)Initialize transmit structures.
*
**********************************************************************/
static void
em_setup_transmit_structures(struct adapter *adapter)
{
struct em_buffer *tx_buffer;
/* Clear the old ring contents */
bzero(adapter->tx_desc_base,
(sizeof(struct e1000_tx_desc)) * adapter->num_tx_desc);
/* Free any existing TX buffers */
for (int i = 0; i < adapter->num_tx_desc; i++, tx_buffer++) {
tx_buffer = &adapter->tx_buffer_area[i];
bus_dmamap_sync(adapter->txtag, tx_buffer->map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(adapter->txtag, tx_buffer->map);
m_freem(tx_buffer->m_head);
tx_buffer->m_head = NULL;
tx_buffer->next_eop = -1;
}
/* Reset state */
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->next_avail_tx_desc = 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
adapter->next_tx_to_clean = 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->num_tx_desc_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
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
bus_dmamap_sync(adapter->txdma.dma_tag, adapter->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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
{
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;
u64 bus_addr;
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");
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Setup the Base and Length of the Tx Descriptor 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
bus_addr = adapter->txdma.dma_paddr;
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_TDLEN(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
adapter->num_tx_desc * sizeof(struct e1000_tx_desc));
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_TDBAH(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
(u32)(bus_addr >> 32));
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_TDBAL(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
(u32)bus_addr);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Setup the HW Tx Head and Tail descriptor pointers */
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_TDT(0), 0);
E1000_WRITE_REG(&adapter->hw, E1000_TDH(0), 0);
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
HW_DEBUGOUT2("Base = %x, Length = %x\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
E1000_READ_REG(&adapter->hw, E1000_TDBAL(0)),
E1000_READ_REG(&adapter->hw, E1000_TDLEN(0)));
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);
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
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 Transmit Descriptor Base Settings */
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
adapter->txd_cmd = E1000_TXD_CMD_IFCS;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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->tx_int_delay.value > 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;
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
}
/*********************************************************************
*
* Free all transmit related data structures.
*
**********************************************************************/
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
{
- 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 em_buffer *tx_buffer;
- 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
- 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("free_transmit_structures: begin");
- 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
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->tx_buffer_area != 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
for (int i = 0; i < adapter->num_tx_desc; i++) {
tx_buffer = &adapter->tx_buffer_area[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
if (tx_buffer->m_head != 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
bus_dmamap_sync(adapter->txtag, tx_buffer->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_POSTWRITE);
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
bus_dmamap_unload(adapter->txtag,
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
tx_buffer->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
m_freem(tx_buffer->m_head);
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
tx_buffer->m_head = 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
} else if (tx_buffer->map != 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
bus_dmamap_unload(adapter->txtag,
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
tx_buffer->map);
if (tx_buffer->map != 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
bus_dmamap_destroy(adapter->txtag,
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
tx_buffer->map);
tx_buffer->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
}
}
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->tx_buffer_area != NULL) {
free(adapter->tx_buffer_area, M_DEVBUF);
adapter->tx_buffer_area = 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
}
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->txtag != NULL) {
bus_dma_tag_destroy(adapter->txtag);
adapter->txtag = 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
}
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
*
**********************************************************************/
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
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_transmit_checksum_setup(struct adapter *adapter, 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
{
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_context_desc *TXD;
- 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;
struct tcp_hdr *th;
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 curr_txd, ehdrlen;
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
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;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
/* Setup checksum offload context. */
curr_txd = adapter->next_avail_tx_desc;
tx_buffer = &adapter->tx_buffer_area[curr_txd];
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
TXD = (struct e1000_context_desc *) &adapter->tx_desc_base[curr_txd];
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.
*/
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;
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
#ifdef EM_TIMESYNC
case ETHERTYPE_IEEE1588:
*txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
break;
#endif
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
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) {
/*
* Start offset for payload checksum calculation.
* End offset for payload checksum calculation.
* Offset of place to put the checksum.
*/
th = (struct tcp_hdr *)(mp->m_data + hdr_len);
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;
*txd_upper |= E1000_TXD_POPTS_TXSM << 8;
}
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
{
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
#ifdef EM_TIMESYNC
void *hdr = (caddr_t) ip + ip_hlen;
struct udphdr *uh = (struct udphdr *)hdr;
if (uh->uh_dport == htons(TSYNC_PORT)) {
*txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
IOCTL_DEBUGOUT("@@@ Sending Event Packet\n");
}
#endif
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (mp->m_pkthdr.csum_flags & CSUM_UDP) {
/*
* Start offset for header checksum calculation.
* End offset for header checksum calculation.
* Offset of place to put the checksum.
*/
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);
*txd_upper |= E1000_TXD_POPTS_TXSM << 8;
}
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
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
#ifdef EM_TIMESYNC
/*
** We might be here just for TIMESYNC
** which means we don't need the context
** descriptor.
*/
if (!mp->m_pkthdr.csum_flags & CSUM_OFFLOAD)
return;
#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
*txd_lower = E1000_TXD_CMD_DEXT | /* Extended descr type */
E1000_TXD_DTYP_D; /* Data descr */
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);
- 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
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 (++curr_txd == adapter->num_tx_desc)
- 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
curr_txd = 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
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_avail--;
adapter->next_avail_tx_desc = curr_txd;
- 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
#if __FreeBSD_version >= 700000
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
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_tso_setup(struct adapter *adapter, struct mbuf *mp, u32 *txd_upper,
u32 *txd_lower)
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
struct e1000_context_desc *TXD;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +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;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
struct ip *ip;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
struct ip6_hdr *ip6;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
struct tcphdr *th;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
int curr_txd, 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;
curr_txd = adapter->next_avail_tx_desc;
tx_buffer = &adapter->tx_buffer_area[curr_txd];
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
TXD = (struct e1000_context_desc *) &adapter->tx_desc_base[curr_txd];
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 */
(isip6 ? 0 : E1000_TXD_CMD_IP) | /* Do IP csum */
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
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
if (++curr_txd == adapter->num_tx_desc)
curr_txd = 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
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
adapter->num_tx_desc_avail--;
adapter->next_avail_tx_desc = curr_txd;
adapter->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
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
#endif /* __FreeBSD_version >= 700000 */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
*
**********************************************************************/
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_txeof(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
{
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
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_TX_LOCK_ASSERT(adapter);
add locking Reviewed by: Prafulla Deuskar <pdeuskar@FreeBSD.ORG> Sponsored by: FreeBSD Foundation
2003-09-23 00:18:25 +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
if (adapter->num_tx_desc_avail == adapter->num_tx_desc)
return;
- 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
num_avail = adapter->num_tx_desc_avail;
first = adapter->next_tx_to_clean;
tx_desc = &adapter->tx_desc_base[first];
tx_buffer = &adapter->tx_buffer_area[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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
eop_desc = &adapter->tx_desc_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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
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;
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++;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_dmamap_sync(adapter->txtag,
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);
bus_dmamap_unload(adapter->txtag,
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;
- 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 = &adapter->tx_buffer_area[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_desc = &adapter->tx_desc_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) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
eop_desc = &adapter->tx_desc_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
}
bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
adapter->next_tx_to_clean = first;
/*
* If we have enough room, clear IFF_DRV_OACTIVE to tell the stack
* that it is OK to send packets.
* If there are no pending descriptors, clear the timeout. Otherwise,
* if some descriptors have been freed, restart the timeout.
*/
if (num_avail > EM_TX_CLEANUP_THRESHOLD) {
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
/* All clean, turn off the timer */
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) {
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
adapter->watchdog_timer = 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
} 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
/* Some cleaned, reset the timer */
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_avail)
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
adapter->watchdog_timer = EM_TX_TIMEOUT;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->num_tx_desc_avail = num_avail;
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;
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
/*********************************************************************
*
* When Link is lost sometimes there is work still in the TX ring
* which will result in a watchdog, rather than allow that do an
* attempted cleanup and then reinit here. Note that this has been
* seens mostly with fiber adapters.
*
**********************************************************************/
static void
em_tx_purge(struct adapter *adapter)
{
if ((!adapter->link_active) && (adapter->watchdog_timer)) {
EM_TX_LOCK(adapter);
em_txeof(adapter);
EM_TX_UNLOCK(adapter);
if (adapter->watchdog_timer) { /* Still not clean? */
adapter->watchdog_timer = 0;
em_init_locked(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
/*********************************************************************
*
* Get a buffer from system mbuf buffer pool.
*
**********************************************************************/
static int
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_get_buf(struct adapter *adapter, int i)
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
{
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;
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 em_buffer *rx_buffer;
int error, nsegs;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
if (m == NULL) {
adapter->mbuf_cluster_failed++;
return (ENOBUFS);
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
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
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
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 <= (MCLBYTES - ETHER_ALIGN))
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
m_adj(m, ETHER_ALIGN);
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
/*
* Using memory from the mbuf cluster pool, invoke the
* bus_dma machinery to arrange the memory mapping.
*/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 = bus_dmamap_load_mbuf_sg(adapter->rxtag,
adapter->rx_sparemap, m, segs, &nsegs, BUS_DMA_NOWAIT);
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 (error != 0) {
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
m_free(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
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
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 nsegs is wrong then the stack is corrupt. */
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
KASSERT(nsegs == 1, ("Too many segments returned!"));
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
rx_buffer = &adapter->rx_buffer_area[i];
if (rx_buffer->m_head != NULL)
bus_dmamap_unload(adapter->rxtag, rx_buffer->map);
map = rx_buffer->map;
rx_buffer->map = adapter->rx_sparemap;
adapter->rx_sparemap = map;
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
bus_dmamap_sync(adapter->rxtag, rx_buffer->map, BUS_DMASYNC_PREREAD);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
rx_buffer->m_head = m;
adapter->rx_desc_base[i].buffer_addr = htole64(segs[0].ds_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
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
}
/*********************************************************************
*
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
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_receive_structures(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
struct em_buffer *rx_buffer;
int i, 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
adapter->rx_buffer_area = 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);
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->rx_buffer_area == 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
#if __FreeBSD_version >= 700000
error = bus_dma_tag_create(bus_get_dma_tag(dev), /* parent */
#else
error = bus_dma_tag_create(NULL, /* parent */
#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
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 */
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->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) */
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
error = bus_dmamap_create(adapter->rxtag, BUS_DMA_NOWAIT,
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->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
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
rx_buffer = adapter->rx_buffer_area;
for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
error = bus_dmamap_create(adapter->rxtag, BUS_DMA_NOWAIT,
- 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
&rx_buffer->map);
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
}
/*********************************************************************
*
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
* (Re)initialize receive structures.
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
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_receive_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
{
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
struct em_buffer *rx_buffer;
int i, error;
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
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 descriptor ring */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
bzero(adapter->rx_desc_base,
(sizeof(struct e1000_rx_desc)) * adapter->num_rx_desc);
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
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
/* Free current RX buffers. */
rx_buffer = adapter->rx_buffer_area;
for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
if (rx_buffer->m_head != NULL) {
bus_dmamap_sync(adapter->rxtag, rx_buffer->map,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(adapter->rxtag, rx_buffer->map);
m_freem(rx_buffer->m_head);
rx_buffer->m_head = NULL;
}
}
/* Allocate new ones. */
for (i = 0; i < adapter->num_rx_desc; i++) {
error = em_get_buf(adapter, i);
if (error)
return (error);
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Setup our descriptor pointers */
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->next_rx_desc_to_check = 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
bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
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
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
}
/*********************************************************************
*
* 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
{
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
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
- 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_receive_unit: 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
*/
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
rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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 (adapter->hw.mac.type >= e1000_82540) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_RADV,
adapter->rx_abs_int_delay.value);
- 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
/*
* Set the interrupt throttling rate. Value is calculated
* as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
*/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_ITR, DEFAULT_ITR);
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
}
- Move 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)
*/
if (adapter->msix)
for (int i = 0; i < 4; i++)
E1000_WRITE_REG(&adapter->hw,
E1000_EITR_82574(i), DEFAULT_ITR);
/* Disable accelerated ackknowledge */
if (adapter->hw.mac.type == e1000_82574)
E1000_WRITE_REG(&adapter->hw,
E1000_RFCTL, E1000_RFCTL_ACK_DIS);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Setup the Base and Length of the Rx Descriptor 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
bus_addr = adapter->rxdma.dma_paddr;
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_RDLEN(0),
adapter->num_rx_desc * sizeof(struct e1000_rx_desc));
E1000_WRITE_REG(&adapter->hw, E1000_RDBAH(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
(u32)(bus_addr >> 32));
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_RDBAL(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
(u32)bus_addr);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Setup the Receive Control Register */
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
rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO |
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
E1000_RCTL_RDMTS_HALF |
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.mc_filter_type << E1000_RCTL_MO_SHIFT);
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
/* Make sure VLAN Filters are off */
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
rctl &= ~E1000_RCTL_VFE;
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
if (e1000_tbi_sbp_enabled_82543(&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
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
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
rctl &= ~E1000_RCTL_SBP;
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
switch (adapter->rx_buffer_len) {
- 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
default:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
case 2048:
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
rctl |= E1000_RCTL_SZ_2048;
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 the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 4096:
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
rctl |= E1000_RCTL_SZ_4096 |
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_RCTL_BSEX | E1000_RCTL_LPE;
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
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 8192:
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
rctl |= E1000_RCTL_SZ_8192 |
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_RCTL_BSEX | E1000_RCTL_LPE;
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 the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 16384:
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
rctl |= E1000_RCTL_SZ_16384 |
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_RCTL_BSEX | E1000_RCTL_LPE;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
break;
}
- 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
if (ifp->if_mtu > ETHERMTU)
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
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
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
rctl &= ~E1000_RCTL_LPE;
- 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
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Enable 82543 Receive Checksum Offload for TCP and UDP */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if ((adapter->hw.mac.type >= e1000_82543) &&
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_capenable & IFCAP_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_READ_REG(&adapter->hw, E1000_RXCSUM);
rxcsum |= (E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
E1000_WRITE_REG(&adapter->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.
*/
if (adapter->hw.mac.type == e1000_82573)
E1000_WRITE_REG(&adapter->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
/* Enable Receives */
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
E1000_WRITE_REG(&adapter->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
/*
* Setup the HW Rx Head and
* Tail Descriptor Pointers
*/
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_RDH(0), 0);
E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), adapter->num_rx_desc - 1);
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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;
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
}
/*********************************************************************
*
* Free receive related data structures.
*
**********************************************************************/
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_receive_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
{
- 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 em_buffer *rx_buffer;
int i;
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("free_receive_structures: begin");
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 (adapter->rx_sparemap) {
bus_dmamap_destroy(adapter->rxtag, adapter->rx_sparemap);
adapter->rx_sparemap = 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
/* Cleanup any existing 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 (adapter->rx_buffer_area != NULL) {
rx_buffer = adapter->rx_buffer_area;
for (i = 0; i < adapter->num_rx_desc; i++, rx_buffer++) {
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 (rx_buffer->m_head != 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
bus_dmamap_sync(adapter->rxtag, rx_buffer->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_POSTREAD);
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
bus_dmamap_unload(adapter->rxtag,
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
rx_buffer->map);
m_freem(rx_buffer->m_head);
rx_buffer->m_head = NULL;
} else if (rx_buffer->map != 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
bus_dmamap_unload(adapter->rxtag,
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
rx_buffer->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 (rx_buffer->map != 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
bus_dmamap_destroy(adapter->rxtag,
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
rx_buffer->map);
rx_buffer->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
}
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 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->rx_buffer_area != NULL) {
free(adapter->rx_buffer_area, M_DEVBUF);
adapter->rx_buffer_area = 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
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 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->rxtag != NULL) {
bus_dma_tag_destroy(adapter->rxtag);
adapter->rxtag = 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
}
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.
*
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
*********************************************************************/
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
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_rxeof(struct adapter *adapter, int count)
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
{
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;;
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 *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
u8 status, accept_frame = 0, eop = 0;
u16 len, desc_len, prev_len_adj;
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
int 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
struct e1000_rx_desc *current_desc;
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 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_RX_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
i = adapter->next_rx_desc_to_check;
current_desc = &adapter->rx_desc_base[i];
bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
Add missing bus_dmamap_sync() calls. If you are using an architecture with a weak memory model or x86 + PAE (or more specifically, your driver is using bounce pages) and you have had problems with em(4), this may fix it. At least this is needed to have em(4) work properly on FreeBSD/arm. Original version by: cognet Reviewed by: tackerman Tested by: cognet
2004-10-19 23:31:44 +00:00
BUS_DMASYNC_POSTREAD);
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 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 (!((current_desc->status) & E1000_RXD_STAT_DD)) {
EM_RX_UNLOCK(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
return (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
}
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
In the em_process_receive_interrupts() cycle check the IFF_DRV_RUNNING flag. This fixes panic, when 'ifconfig em0 down' was called and it calls em_stop() while the em_process_receive_interrupts() has temporarily dropped the lock.
2005-10-19 13:34:48 +00:00
while ((current_desc->status & E1000_RXD_STAT_DD) &&
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
(count != 0) &&
(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
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
struct mbuf *m = NULL;
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
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
mp = adapter->rx_buffer_area[i].m_head;
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
/*
* Can't defer bus_dmamap_sync(9) because TBI_ACCEPT
* needs to access the last received byte in the 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
bus_dmamap_sync(adapter->rxtag, adapter->rx_buffer_area[i].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_POSTREAD);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
accept_frame = 1;
- 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
prev_len_adj = 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
desc_len = le16toh(current_desc->length);
Don't update Rx descriptor status in two different functions. Suggested by: pdeuskar Reviewed by: pdeuskar
2006-08-16 23:55:34 +00:00
status = current_desc->status;
if (status & E1000_RXD_STAT_EOP) {
Add polling support to the "em" driver. Reviewed by: pdeuskar (maintainer) Approved by: re
2002-11-25 19:56:16 +00:00
count--;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
eop = 1;
- 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
if (desc_len < ETHER_CRC_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
len = 0;
prev_len_adj = ETHER_CRC_LEN - desc_len;
} else
len = desc_len - ETHER_CRC_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 {
eop = 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
len = desc_len;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
if (current_desc->errors & E1000_RXD_ERR_FRAME_ERR_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
u8 last_byte;
u32 pkt_len = desc_len;
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->fmp != NULL)
pkt_len += adapter->fmp->m_pkthdr.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
last_byte = *(mtod(mp, caddr_t) + desc_len - 1);
Don't update Rx descriptor status in two different functions. Suggested by: pdeuskar Reviewed by: pdeuskar
2006-08-16 23:55:34 +00:00
if (TBI_ACCEPT(&adapter->hw, 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
current_desc->errors, pkt_len, last_byte,
adapter->min_frame_size, 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
e1000_tbi_adjust_stats_82543(&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
&adapter->stats, pkt_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
adapter->hw.mac.addr,
adapter->max_frame_size);
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 (len > 0)
len--;
} else
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
accept_frame = 0;
}
if (accept_frame) {
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (em_get_buf(adapter, i) != 0) {
ifp->if_iqdrops++;
goto discard;
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 */
mp->m_len = 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
if (adapter->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;
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->fmp = mp; /* Store the first mbuf */
adapter->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;
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
/*
* Adjust length of previous mbuf in chain if
* we received less than 4 bytes in the last
* descriptor.
*/
- 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
if (prev_len_adj > 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->lmp->m_len -= prev_len_adj;
adapter->fmp->m_pkthdr.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
prev_len_adj;
- 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
}
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->lmp->m_next = mp;
adapter->lmp = adapter->lmp->m_next;
adapter->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) {
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->fmp->m_pkthdr.rcvif = ifp;
Cleanup from __FreeBSD_version.
2005-10-14 10:34:46 +00:00
ifp->if_ipackets++;
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_receive_checksum(adapter, current_desc,
adapter->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) &&
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_fixup_rx(adapter) != 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) {
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
VLAN_INPUT_TAG_NEW(ifp, adapter->fmp,
(le16toh(current_desc->special) &
E1000_RXD_SPC_VLAN_MASK));
#else
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
adapter->fmp->m_pkthdr.ether_vtag =
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
(le16toh(current_desc->special) &
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
E1000_RXD_SPC_VLAN_MASK);
adapter->fmp->m_flags |= M_VLANTAG;
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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
skip:
#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
m = adapter->fmp;
adapter->fmp = NULL;
adapter->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++;
discard:
/* Reuse loaded DMA map and just update mbuf chain */
mp = adapter->rx_buffer_area[i].m_head;
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);
if (adapter->fmp != 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
m_freem(adapter->fmp);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
adapter->fmp = NULL;
adapter->lmp = NULL;
}
m = 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. */
current_desc->status = 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
bus_dmamap_sync(adapter->rxdma.dma_tag, adapter->rxdma.dma_map,
- Use BUS_DMASYNC_PREWRITE in em_get_buf(), as the adapter is about to read the descriptors set. - In em_process_receive_interrupts(), call bus_dmamap_sync() for the descriptors set each time we modify one descriptor, instead of doing it only at the function exit, to make sure the adapters know he can re-use the descriptor. This helps on arm with write-back data cache (and possibly on other arches with bounce pages, I don't know) under heavy network load. Without this, if we attempt to process more than num_rx_desc descriptors, the adapter would just stop processing rx interrupts.
2005-10-18 00:42:10 +00:00
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
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;
if (m != 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
adapter->next_rx_desc_to_check = i;
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
/* Unlock for call into stack */
EM_RX_UNLOCK(adapter);
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
(*ifp->if_input)(ifp, m);
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_RX_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
i = adapter->next_rx_desc_to_check;
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
}
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
current_desc = &adapter->rx_desc_base[i];
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
adapter->next_rx_desc_to_check = i;
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
/* Advance the E1000's Receive Queue #0 "Tail Pointer". */
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 (--i < 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
i = adapter->num_rx_desc - 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_RDT(0), i);
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_RX_UNLOCK(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
if (!((current_desc->status) & E1000_RXD_STAT_DD))
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
return (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
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
return (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 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
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_fixup_rx(struct adapter *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;
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
m = adapter->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;
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->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++;
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
m_freem(adapter->fmp);
adapter->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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_receive_checksum(struct adapter *adapter,
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
{
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* 82543 or newer only */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
if ((adapter->hw.mac.type < e1000_82543) ||
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Ignore Checksum bit is set */
(rx_desc->status & E1000_RXD_STAT_IXSM)) {
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
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
em_register_vlan(void *unused, 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;
u32 ctrl, rctl, index, vfta;
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
ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
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
ctrl |= E1000_CTRL_VME;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_CTRL, ctrl);
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
/* Setup for Hardware Filter */
rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
rctl |= E1000_RCTL_VFE;
rctl &= ~E1000_RCTL_CFIEN;
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
/* Make entry in the hardware filter table */
index = ((vtag >> 5) & 0x7F);
vfta = E1000_READ_REG_ARRAY(&adapter->hw, E1000_VFTA, index);
vfta |= (1 << (vtag & 0x1F));
E1000_WRITE_REG_ARRAY(&adapter->hw, E1000_VFTA, index, vfta);
/* Update the frame size */
E1000_WRITE_REG(&adapter->hw, E1000_RLPML,
adapter->max_frame_size + VLAN_TAG_SIZE);
}
/*
* This routine is run via an vlan
* unconfig EVENT
*/
static void
em_unregister_vlan(void *unused, struct ifnet *ifp, u16 vtag)
{
struct adapter *adapter = ifp->if_softc;
u32 index, vfta;
/* Remove entry in the hardware filter table */
index = ((vtag >> 5) & 0x7F);
vfta = E1000_READ_REG_ARRAY(&adapter->hw, E1000_VFTA, index);
vfta &= ~(1 << (vtag & 0x1F));
E1000_WRITE_REG_ARRAY(&adapter->hw, E1000_VFTA, index, vfta);
/* Have all vlans unregistered? */
if (adapter->ifp->if_vlantrunk == NULL) {
u32 rctl;
/* Turn off the filter table */
rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
rctl &= ~E1000_RCTL_VFE;
rctl |= E1000_RCTL_CFIEN;
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
/* Reset the frame size */
E1000_WRITE_REG(&adapter->hw, E1000_RLPML,
adapter->max_frame_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;
if (adapter->msix) {
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;
if (adapter->msix)
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 */
if (adapter->hw.mac.type >= e1000_82571) {
manc |= E1000_MANC_EN_MNG2HOST;
#define E1000_MNG2HOST_PORT_623 (1 << 5)
#define E1000_MNG2HOST_PORT_664 (1 << 6)
manc2h |= E1000_MNG2HOST_PORT_623;
manc2h |= E1000_MNG2HOST_PORT_664;
E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
}
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;
if (adapter->hw.mac.type >= e1000_82571)
manc &= ~E1000_MANC_EN_MNG2HOST;
E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
}
}
/*
* em_get_hw_control sets {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 (only with 82573)
* of the f/w this means that the network i/f is open.
*
*/
static void
em_get_hw_control(struct adapter *adapter)
{
u32 ctrl_ext, swsm;
/* Let firmware know the driver has taken over */
switch (adapter->hw.mac.type) {
case e1000_82573:
swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
swsm | E1000_SWSM_DRV_LOAD);
break;
case e1000_82571:
case e1000_82572:
case e1000_80003es2lan:
case e1000_ich8lan:
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:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
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);
break;
default:
break;
}
}
/*
* em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
* For ASF and Pass Through versions of f/w this means that the
* driver is no longer loaded. For AMT version (only with 82573) i
* of the f/w this means that the network i/f is closed.
*
*/
static void
em_release_hw_control(struct adapter *adapter)
{
u32 ctrl_ext, swsm;
/* Let firmware taken over control of h/w */
switch (adapter->hw.mac.type) {
case e1000_82573:
swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
swsm & ~E1000_SWSM_DRV_LOAD);
break;
case e1000_82571:
case e1000_82572:
case e1000_80003es2lan:
case e1000_ich8lan:
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:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
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);
break;
default:
break;
}
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
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
*/
void
em_enable_wakeup(device_t dev)
{
u16 cap, status;
u8 id;
/* First find the capabilities pointer*/
cap = pci_read_config(dev, PCIR_CAP_PTR, 2);
/* Read the PM Capabilities */
id = pci_read_config(dev, cap, 1);
if (id != PCIY_PMG) /* Something wrong */
return;
/* OK, we have the power capabilities, so
now get the status register */
cap += PCIR_POWER_STATUS;
status = pci_read_config(dev, cap, 2);
status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
pci_write_config(dev, cap, status, 2);
return;
}
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
/*********************************************************************
* 82544 Coexistence issue workaround.
* There are 2 issues.
* 1. Transmit Hang issue.
* To detect this issue, following equation can be used...
- 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
* SIZE[3:0] + ADDR[2:0] = SUM[3:0].
* If SUM[3:0] is in between 1 to 4, we will have this issue.
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
*
* 2. DAC issue.
* To detect this issue, following equation can be used...
- 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
* SIZE[3:0] + ADDR[2:0] = SUM[3:0].
* If SUM[3:0] is in between 9 to c, we will have this issue.
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
*
*
* WORKAROUND:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 do not have ending address
* as 1,2,3,4(Hang) or 9,a,b,c (DAC)
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
*
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
static u32
em_fill_descriptors (bus_addr_t address, u32 length,
- 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
PDESC_ARRAY desc_array)
{
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 safe_terminator;
- 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
/* Since issue is sensitive to length and address.*/
/* Let us first check the address...*/
if (length <= 4) {
desc_array->descriptor[0].address = address;
desc_array->descriptor[0].length = length;
desc_array->elements = 1;
return (desc_array->elements);
}
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
safe_terminator = (u32)((((u32)address & 0x7) +
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
(length & 0xF)) & 0xF);
- 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 it does not fall between 0x1 to 0x4 and 0x9 to 0xC then return */
if (safe_terminator == 0 ||
(safe_terminator > 4 &&
safe_terminator < 9) ||
(safe_terminator > 0xC &&
safe_terminator <= 0xF)) {
desc_array->descriptor[0].address = address;
desc_array->descriptor[0].length = length;
desc_array->elements = 1;
return (desc_array->elements);
}
desc_array->descriptor[0].address = address;
desc_array->descriptor[0].length = length - 4;
desc_array->descriptor[1].address = address + (length - 4);
desc_array->descriptor[1].length = 4;
desc_array->elements = 2;
return (desc_array->elements);
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
}
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;
- 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));
- 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, "fifo workaround = %lld, fifo_reset_count = %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->tx_fifo_wrk_cnt,
(long long)adapter->tx_fifo_reset_cnt);
- 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, "hw tdh = %d, hw tdt = %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
E1000_READ_REG(&adapter->hw, E1000_TDH(0)),
E1000_READ_REG(&adapter->hw, E1000_TDT(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, "hw rdh = %d, hw rdt = %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
E1000_READ_REG(&adapter->hw, E1000_RDH(0)),
E1000_READ_REG(&adapter->hw, E1000_RDT(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, "Num Tx descriptors avail = %d\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->num_tx_desc_avail);
- 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 Descriptors not avail1 = %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->no_tx_desc_avail1);
- 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 Descriptors not avail2 = %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->no_tx_desc_avail2);
- 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);
device_printf(dev, "Driver tx dma failure in encap = %ld\n",
adapter->no_tx_dma_setup);
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);
device_printf(dev, "RX overruns = %ld\n", adapter->rx_overruns);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
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
device_printf(dev, "RX MSIX IRQ = %ld TX MSIX IRQ = %ld"
" LINK MSIX IRQ = %ld\n", adapter->rx_irq,
adapter->tx_irq , adapter->link_irq);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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, "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;
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
int error;
int usecs;
int ticks;
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
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
#ifndef EM_LEGACY_IRQ
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);
}
#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
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
#ifdef EM_TIMESYNC
/*
* Initialize the Time Sync Feature
*/
static int
em_tsync_init(struct adapter *adapter)
{
device_t dev = adapter->dev;
u32 tx_ctl, rx_ctl;
E1000_WRITE_REG(&adapter->hw, E1000_TIMINCA, (1<<24) |
20833/PICOSECS_PER_TICK);
adapter->last_stamp = E1000_READ_REG(&adapter->hw, E1000_SYSTIML);
adapter->last_stamp |= (u64)E1000_READ_REG(&adapter->hw,
E1000_SYSTIMH) << 32ULL;
/* Enable the TX side */
tx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCTXCTL);
tx_ctl |= 0x10;
E1000_WRITE_REG(&adapter->hw, E1000_TSYNCTXCTL, tx_ctl);
E1000_WRITE_FLUSH(&adapter->hw);
tx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCTXCTL);
if ((tx_ctl & 0x10) == 0) {
device_printf(dev, "Failed to enable TX timestamping\n");
return (ENXIO);
}
/* Enable RX */
rx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCRXCTL);
rx_ctl |= 0x10; /* Enable the feature */
rx_ctl |= 0x0a; /* This value turns on Ver 1 and 2 */
E1000_WRITE_REG(&adapter->hw, E1000_TSYNCRXCTL, rx_ctl);
/*
* Ethertype Stamping (Ethertype = 0x88F7)
*/
E1000_WRITE_REG(&adapter->hw, E1000_RXMTRL, htonl(0x440088f7));
/*
* Source Port Queue Filter Setup:
* this is for UDP port filtering
*/
E1000_WRITE_REG(&adapter->hw, E1000_RXUDP, htons(TSYNC_PORT));
/* Protocol = UDP, enable Timestamp, and filter on source/protocol */
E1000_WRITE_FLUSH(&adapter->hw);
rx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCRXCTL);
if ((rx_ctl & 0x10) == 0) {
device_printf(dev, "Failed to enable RX timestamping\n");
return (ENXIO);
}
device_printf(dev, "IEEE 1588 Precision Time Protocol enabled\n");
return (0);
}
/*
* Disable the Time Sync Feature
*/
static void
em_tsync_disable(struct adapter *adapter)
{
u32 tx_ctl, rx_ctl;
tx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCTXCTL);
tx_ctl &= ~0x10;
E1000_WRITE_REG(&adapter->hw, E1000_TSYNCTXCTL, tx_ctl);
E1000_WRITE_FLUSH(&adapter->hw);
/* Invalidate TX Timestamp */
E1000_READ_REG(&adapter->hw, E1000_TXSTMPH);
tx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCTXCTL);
if (tx_ctl & 0x10)
HW_DEBUGOUT("Failed to disable TX timestamping\n");
rx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCRXCTL);
rx_ctl &= ~0x10;
E1000_WRITE_REG(&adapter->hw, E1000_TSYNCRXCTL, rx_ctl);
E1000_WRITE_FLUSH(&adapter->hw);
/* Invalidate RX Timestamp */
E1000_READ_REG(&adapter->hw, E1000_RXSATRH);
rx_ctl = E1000_READ_REG(&adapter->hw, E1000_TSYNCRXCTL);
if (rx_ctl & 0x10)
HW_DEBUGOUT("Failed to disable RX timestamping\n");
return;
}
#endif /* EM_TIMESYNC */
Reference in New Issue Copy Permalink