d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
2fdb65fd82
freebsd-nq/sys/dev/em/if_em.c

5020 lines
146 KiB
C
Raw Normal View History

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
Copyright (c) 2001-2007, Intel Corporation
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
All rights reserved.
Corrected license in the source files. It should say "MUST" instead of "MAY". MFC after: 2 days
2002-09-24 16:27:59 +00:00
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
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
Corrected license in the source files. It should say "MUST" instead of "MAY". MFC after: 2 days
2002-09-24 16:27:59 +00:00
1. Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
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
Corrected license in the source files. It should say "MUST" instead of "MAY". MFC after: 2 days
2002-09-24 16:27:59 +00:00
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
This 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
3. Neither the name of the Intel Corporation nor the names of its
Corrected license in the source files. It should say "MUST" instead of "MAY". MFC after: 2 days
2002-09-24 16:27:59 +00:00
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
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 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
Corrected license in the source files. It should say "MUST" instead of "MAY". MFC after: 2 days
2002-09-24 16:27:59 +00:00
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.
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
***************************************************************************/
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
/*$FreeBSD$*/
- 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>
#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"
#include "e1000_82575.h"
#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 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
char em_driver_version[] = "Version - 6.5.2";
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},
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},
- Second style(9) 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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},
{ 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},
{ 0x8086, E1000_DEV_ID_82575EB_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES,
PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82575EM_COPPER, PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82575EM_FIBER_SERDES,
PCI_ANY_ID, PCI_ANY_ID, 0},
{ 0x8086, E1000_DEV_ID_82575GB_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
/* 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 *);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
static int em_allocate_intr(struct adapter *);
static void em_free_intr(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 *);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_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 *);
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 *,
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
uint32_t *, uint32_t *);
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
static boolean_t em_tx_adv_ctx_setup(struct adapter *, struct mbuf *);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 boolean_t em_tso_setup(struct adapter *, struct mbuf *, uint32_t *,
uint32_t *);
static boolean_t em_tso_adv_setup(struct adapter *, struct mbuf *, uint32_t *);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
static void em_enable_vlans(struct adapter *);
static int em_encap(struct adapter *, struct mbuf **);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_adv_encap(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 *);
- Second style(9) 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_is_valid_ether_addr(uint8_t *);
static int em_sysctl_stats(SYSCTL_HANDLER_ARGS);
static int em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
static uint32_t em_fill_descriptors (bus_addr_t address, uint32_t length,
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
#ifdef DEVICE_POLLING
static poll_handler_t em_poll;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
static void em_intr(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 em_intr_fast(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);
static void em_handle_link(void *context, int pending);
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
#endif
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
/*********************************************************************
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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;
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);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
#ifndef DEVICE_POLLING
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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];
uint16_t pci_vendor_id = 0;
uint16_t pci_device_id = 0;
uint16_t pci_subvendor_id = 0;
uint16_t 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;
EM_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)),
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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, "debug_info", 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
- 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_init_mtx(&adapter->timer, &adapter->mtx, 0);
callout_init_mtx(&adapter->tx_fifo_timer, &adapter->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
/* Determine hardware revision */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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) ||
(adapter->hw.mac.type == e1000_ich9lan)) {
int rid = EM_BAR_TYPE_FLASH;
adapter->flash_mem = bus_alloc_resource_any(dev,
SYS_RES_MEMORY, &rid, RF_ACTIVE);
/* This is used in the shared code */
adapter->hw.flash_address = (u8 *)adapter->flash_mem;
adapter->osdep.flash_bus_space_tag =
rman_get_bustag(adapter->flash_mem);
adapter->osdep.flash_bus_space_handle =
rman_get_bushandle(adapter->flash_mem);
}
/* 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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
#ifndef DEVICE_POLLING
/* 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;
adapter->hw.phy.wait_for_link = FALSE;
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 */
if (adapter->hw.media_type == e1000_media_type_copper) {
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
/*
* Set the max frame size assuming standard ethernet
* sized frames.
*/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.max_frame_size =
ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.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;
}
}
if (e1000_read_part_num(&adapter->hw, &(adapter->part_num)) < 0) {
device_printf(dev, "EEPROM read error "
"reading part number\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
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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_allocate_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
/* 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Get control from any management/hw control */
if (((adapter->hw.mac.type != e1000_82573) &&
(adapter->hw.mac.type != e1000_ich8lan) &&
(adapter->hw.mac.type != e1000_ich9lan)) ||
!e1000_check_mng_mode(&adapter->hw))
em_get_hw_control(adapter);
/* 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);
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);
e1000_remove_device(&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
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:
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_free_intr(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_free_pci_resources(adapter);
EM_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");
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_disable_intr(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_free_intr(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_LOCK(adapter);
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);
if (((adapter->hw.mac.type != e1000_82573) &&
(adapter->hw.mac.type != e1000_ich8lan) &&
(adapter->hw.mac.type != e1000_ich9lan)) ||
!e1000_check_mng_mode(&adapter->hw))
em_release_hw_control(adapter);
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);
}
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_UNLOCK(adapter);
ether_ifdetach(adapter->ifp);
Cleanup from __FreeBSD_version.
2005-10-14 10:34:46 +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
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_remove_device(&adapter->hw);
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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
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
em_release_manageability(adapter);
if (((adapter->hw.mac.type != e1000_82573) &&
(adapter->hw.mac.type != e1000_ich8lan) &&
(adapter->hw.mac.type != e1000_ich9lan)) ||
!e1000_check_mng_mode(&adapter->hw))
em_release_hw_control(adapter);
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);
}
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
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
/* Get control from any management/hw control */
if (((adapter->hw.mac.type != e1000_82573) &&
(adapter->hw.mac.type != e1000_ich8lan) &&
(adapter->hw.mac.type != e1000_ich9lan)) ||
!e1000_check_mng_mode(&adapter->hw))
em_get_hw_control(adapter);
em_init_manageability(adapter);
Tidy up em_resume(): - Don't call em_init_locked() twice. - Collapse two if() blocks into one.
2005-12-28 08:58:28 +00:00
if ((ifp->if_flags & IFF_UP) &&
(ifp->if_drv_flags & IFF_DRV_RUNNING))
A style nit.
2005-12-28 09:37:04 +00:00
em_start_locked(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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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_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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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.
*
* We now use a pointer to accomodate legacy and
* advanced transmit functions.
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
if (adapter->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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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;
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;
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:
case SIOCGIFADDR:
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)) {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
em_init_locked(adapter);
EM_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
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;
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
uint16_t 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)");
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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:
case e1000_82575:
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
/* ICH8 does not support jumbo frames */
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) {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.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);
EM_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)");
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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) &
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
IFF_PROMISC) {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_stop(adapter);
adapter->if_flags = ifp->if_flags;
EM_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) {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
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);
EM_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 */
EM_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)) {
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
EM_UNLOCK(adapter);
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;
}
EM_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);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
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;
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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 */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
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;
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_UNLOCK(adapter);
Big polling(4) cleanup. o Axe poll in trap. o Axe IFF_POLLING flag from if_flags. o Rework revision 1.21 (Giant removal), in such a way that poll_mtx is not dropped during call to polling handler. This fixes problem with idle polling. o Make registration and deregistration from polling in a functional way, insted of next tick/interrupt. o Obsolete kern.polling.enable. Polling is turned on/off with ifconfig. Detailed kern_poll.c changes: - Remove polling handler flags, introduced in 1.21. The are not needed now. - Forget and do not check if_flags, if_capenable and if_drv_flags. - Call all registered polling handlers unconditionally. - Do not drop poll_mtx, when entering polling handlers. - In ether_poll() NET_LOCK_GIANT prior to locking poll_mtx. - In netisr_poll() axe the block, where polling code asks drivers to unregister. - In netisr_poll() and ether_poll() do polling always, if any handlers are present. - In ether_poll_[de]register() remove a lot of error hiding code. Assert that arguments are correct, instead. - In ether_poll_[de]register() use standard return values in case of error or success. - Introduce poll_switch() that is a sysctl handler for kern.polling.enable. poll_switch() goes through interface list and enabled/disables polling. A message that kern.polling.enable is deprecated is printed. Detailed driver changes: - On attach driver announces IFCAP_POLLING in if_capabilities, but not in if_capenable. - On detach driver calls ether_poll_deregister() if polling is enabled. - In polling handler driver obtains its lock and checks IFF_DRV_RUNNING flag. If there is no, then unlocks and returns. - In ioctl handler driver checks for IFCAP_POLLING flag requested to be set or cleared. Driver first calls ether_poll_[de]register(), then obtains driver lock and [dis/en]ables interrupts. - In interrupt handler driver checks IFCAP_POLLING flag in if_capenable. If present, then returns.This is important to protect from spurious interrupts. Reviewed by: ru, sam, jhb
2005-10-01 18:56:19 +00:00
}
}
#endif
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
if (mask & IFCAP_HWCSUM) {
Implemented per-interface polling(4) control.
2004-04-11 21:01:12 +00:00
ifp->if_capenable ^= IFCAP_HWCSUM;
Implement support for controlling VLAN_HWTAGGING through ioctl(SIOCSIFCAP). This includes not only toggling the flag in if_capenable, but also really reconfiguring the hardware. Approved by: tackerman (as the em(4) maintainer)
2004-06-04 16:57:18 +00:00
reinit = 1;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (mask & IFCAP_TSO4) {
ifp->if_capenable ^= IFCAP_TSO4;
reinit = 1;
}
Implement support for controlling VLAN_HWTAGGING through ioctl(SIOCSIFCAP). This includes not only toggling the flag in if_capenable, but also really reconfiguring the hardware. Approved by: tackerman (as the em(4) maintainer)
2004-06-04 16:57:18 +00:00
if (mask & IFCAP_VLAN_HWTAGGING) {
ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
reinit = 1;
}
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (reinit && (ifp->if_drv_flags & IFF_DRV_RUNNING))
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_init(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
VLAN_CAPABILITIES(ifp);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
break;
Merge in new driver version from Intel - 3.2.18. The most important change is support for adapters based on 82571 and 82572 chips. Tested on: 82547EI on i386 Tested on: 82540EM on sparc64
2005-11-24 01:44:49 +00:00
}
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
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_LOCK_ASSERT(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
** The timer is set to 5 every time start queues a packet.
** Then txeof keeps resetting to 5 as long as it cleans at
** least one descriptor.
** Finally, anytime all descriptors are clean the timer is
** set to 0.
*/
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 (adapter->watchdog_timer == 0 || --adapter->watchdog_timer)
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
return;
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;
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++;
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;
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
uint32_t 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");
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_stop(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
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 */
if (adapter->hw.mac.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_82575:
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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:
#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. */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.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);
/*
* With 82571 controllers, LAA may be overwritten
* due to controller reset from the other port.
*/
if (adapter->hw.mac.type == e1000_82571)
e1000_set_laa_state_82571(&adapter->hw, TRUE);
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
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 (ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_vlans(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 the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
if (ifp->if_capenable & IFCAP_TSO4)
ifp->if_hwassist |= CSUM_TSO;
Move the initialization of the hardware capabilities in em_init_locked() before em_setup_transmit_structures() as it needs this information to properly set up TSO parameters. Reviewed by: jfv
2006-09-21 19:14:51 +00:00
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Configure for OS presence */
em_init_manageability(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Prepare transmit descriptors and buffers */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_setup_transmit_structures(adapter)) {
device_printf(dev, "Could not setup transmit structures\n");
em_stop(adapter);
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_transmit_unit(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Setup Multicast table */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_set_multi(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
/* Prepare receive descriptors and buffers */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
if (em_setup_receive_structures(adapter)) {
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
device_printf(dev, "Could not setup receive structures\n");
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_stop(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
return;
}
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_initialize_receive_unit(adapter);
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
Commit the results of the typo hunt by Darren Pilgrim. This change affects documentation and comments only, no real code involved. PR: misc/101245 Submitted by: Darren Pilgrim <darren pilgrim bitfreak org> Tested by: md5(1) MFC after: 1 week
2006-08-04 07:56:35 +00:00
/* Don't lose promiscuous settings */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
em_set_promisc(adapter);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. This makes it easier for us to get the changes into -current and to -stable quickly.
2006-08-03 19:05:04 +00:00
callout_reset(&adapter->timer, hz, em_local_timer, adapter);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
e1000_clear_hw_cntrs_base_generic(&adapter->hw);
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
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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
em_init_locked(adapter);
EM_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;
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
uint32_t 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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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) {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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;
e1000_check_for_link(&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
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
}
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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, count);
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);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_UNLOCK(adapter);
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
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;
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
struct ifnet *ifp;
uint32_t reg_icr;
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
EM_LOCK(adapter);
ifp = adapter->ifp;
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_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_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;
}
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;
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);
em_txeof(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;
e1000_check_for_link(&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
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);
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
}
if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
em_start_locked(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
EM_UNLOCK(adapter);
Add polling support to the "em" driver. Reviewed by: pdeuskar (maintainer) Approved by: re
2002-11-25 19:56:16 +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
#else /* if not DEVICE_POLLING, 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;
struct ifnet *ifp;
ifp = adapter->ifp;
EM_LOCK(adapter);
- 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
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
EM_UNLOCK(adapter);
return;
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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;
e1000_check_for_link(&adapter->hw);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_update_link_status(adapter);
callout_reset(&adapter->timer, hz, em_local_timer, adapter);
EM_UNLOCK(adapter);
}
static void
em_handle_rxtx(void *context, int pending)
{
struct adapter *adapter = context;
struct ifnet *ifp;
NET_LOCK_GIANT();
ifp = adapter->ifp;
/*
* TODO:
* It should be possible to run the tx clean loop without the lock.
*/
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
if (em_rxeof(adapter, adapter->rx_process_limit) != 0)
taskqueue_enqueue(adapter->tq, &adapter->rxtx_task);
EM_LOCK(adapter);
em_txeof(adapter);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
em_start_locked(ifp);
EM_UNLOCK(adapter);
}
em_enable_intr(adapter);
NET_UNLOCK_GIANT();
}
/*********************************************************************
*
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
* Fast Interrupt Service routine
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
*
*********************************************************************/
o break newbus api: add a new argument of type driver_filter_t to bus_setup_intr() o add an int return code to all fast handlers o retire INTR_FAST/IH_FAST For more info: http://docs.freebsd.org/cgi/getmsg.cgi?fetch=465712+0+current/freebsd-current Reviewed by: many Approved by: re@
2007-02-23 12:19:07 +00:00
static int
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_intr_fast(void *arg)
{
struct adapter *adapter = arg;
struct ifnet *ifp;
uint32_t reg_icr;
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)
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
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)
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
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)
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
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 */
if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
taskqueue_enqueue(taskqueue_fast, &adapter->link_task);
if (reg_icr & E1000_ICR_RXO)
adapter->rx_overruns++;
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
return (FILTER_HANDLED);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
}
#endif /* ! DEVICE_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
/*********************************************************************
*
* 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");
Add locking to the ifmedia callouts. Reviewed by: jvogel, yongari
2006-09-01 16:08:36 +00:00
EM_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
e1000_check_for_link(&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
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) {
EM_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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.media_type == e1000_media_type_fiber) ||
(adapter->hw.media_type == e1000_media_type_internal_serdes)) {
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;
}
Add locking to the ifmedia callouts. Reviewed by: jvogel, yongari
2006-09-01 16:08:36 +00:00
EM_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
Add locking to the ifmedia callouts. Reviewed by: jvogel, yongari
2006-09-01 16:08:36 +00:00
EM_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);
EM_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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_encap(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;
uint32_t 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;
- 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
Correct a bug in the if_em driver relating to the use of vlans with promiscuous mode introduced in 1.45, which programs the em card not to strip or prepend tags when in promiscuous mode without also modifying behavior to manually prepend a vlan header in the event that the card isn't doing it on transmit. Due to a feature of card operation, if the global VLAN prepend/strip register isn't set, setting the VLAN tag flag on individual packet descriptors will cause the packet to be transmitted using ISL encapsulation rather than 802.1Q VLAN encapsulation. This fix causes em_encap() to prepend the header by tracking whether the card is configured to temporarily disable prepending/stripping due to promiscuous mode. As a result, entering promiscuous mode on the parent em interface no longer causes vlans to appear to "wedge" or transmit ISL-encapsulated frames, which typically will not be configured/spoken by the other endpoints on the VLAN trunk. This bug may also exist in other drivers, and the additional vlan encapsulation logic should be abstracted and centralized in if_vlan.c if so. RELENG_5_3 candidate. MFC after: 1 week Tested by: pjd, rwatson Reported by: astesin at ukrtelecom dot net Reported by: Mike Tancsa <mike at sentex dot net> Reported by: Iasen Kostov <tbyte at OTEL dot net>
2004-11-12 11:03:07 +00:00
m_head = *m_headp;
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +00:00
txd_upper = txd_lower = txd_used = txd_saved = 0;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
do_tso = ((m_head->m_pkthdr.csum_flags & CSUM_TSO) != 0);
Correct a bug in the if_em driver relating to the use of vlans with promiscuous mode introduced in 1.45, which programs the em card not to strip or prepend tags when in promiscuous mode without also modifying behavior to manually prepend a vlan header in the event that the card isn't doing it on transmit. Due to a feature of card operation, if the global VLAN prepend/strip register isn't set, setting the VLAN tag flag on individual packet descriptors will cause the packet to be transmitted using ISL encapsulation rather than 802.1Q VLAN encapsulation. This fix causes em_encap() to prepend the header by tracking whether the card is configured to temporarily disable prepending/stripping due to promiscuous mode. As a result, entering promiscuous mode on the parent em interface no longer causes vlans to appear to "wedge" or transmit ISL-encapsulated frames, which typically will not be configured/spoken by the other endpoints on the VLAN trunk. This bug may also exist in other drivers, and the additional vlan encapsulation logic should be abstracted and centralized in if_vlan.c if so. RELENG_5_3 candidate. MFC after: 1 week Tested by: pjd, rwatson Reported by: astesin at ukrtelecom dot net Reported by: Mike Tancsa <mike at sentex dot net> Reported by: Iasen Kostov <tbyte at OTEL dot net>
2004-11-12 11:03:07 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 */
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 (em_tso_setup(adapter, m_head, &txd_upper, &txd_lower))
/* we need to make a final sentinel transmit desc */
tso_desc = TRUE;
else if (m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD)
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;
uint32_t array_elements, counter;
/*
* 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(
(adapter->txd_cmd | txd_lower | (uint16_t)
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 {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 = 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);
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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (m_head->m_flags & M_VLANTAG) {
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
/* Set the vlan id. */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
ctxd->upper.fields.special =
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
htole16(m_head->m_pkthdr.ether_vtag);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
/* Tell hardware to add tag */
ctxd->lower.data |= htole32(E1000_TXD_CMD_VLE);
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
tx_buffer->m_head = m_head;
Rework the transmit register handling. In em_encap() store index of the EOP descriptor in the first descriptor of the packet. And then in em_txeof() search for DD bits set only in the EOP descriptors, embedding the cleanup of all packet's descriptors into inner loop. This change is important for future chips, where DD bit is going to be set only on the EOP descriptors. Submitted by: jfv
2006-10-31 16:19:21 +00:00
tx_buffer_mapped->map = tx_buffer->map;
Make sure to use the same DMA map in DMA map load/unload operations by remembering a map used in bus_dmamap_load_mbuf_sg(9). I have no idea how it could ever worked before. This fixes a warning generated by a diagnostic check in sun4v iommu driver. Reported by: jb Tested by: jb(sun4v)
2006-07-27 00:26:20 +00:00
tx_buffer->map = map;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_TDT, i);
if (adapter->hw.mac.type == e1000_82547)
em_82547_update_fifo_head(adapter,
m_head->m_pkthdr.len);
}
return (0);
}
/*********************************************************************
*
* This routine maps the mbufs to Advanced TX descriptors.
* used by the 82575 adapter. It also needs no workarounds.
*
**********************************************************************/
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_adv_encap(struct adapter *adapter, struct mbuf **m_headp)
{
bus_dma_segment_t segs[EM_MAX_SCATTER];
bus_dmamap_t map;
struct em_buffer *tx_buffer, *tx_buffer_mapped;
union e1000_adv_tx_desc *txd = NULL;
struct mbuf *m_head;
u32 olinfo_status = 0, cmd_type_len = 0;
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
u32 paylen = 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
int nsegs, i, j, error, first, last = 0;
m_head = *m_headp;
/* Set basic descriptor constants */
cmd_type_len |= E1000_ADVTXD_DTYP_DATA;
cmd_type_len |= E1000_ADVTXD_DCMD_IFCS | E1000_ADVTXD_DCMD_DEXT;
/*
* Force a cleanup if number of TX descriptors
* available hits the threshold
*/
if (adapter->num_tx_desc_avail <= EM_TX_CLEANUP_THRESHOLD) {
em_txeof(adapter);
/* Now do we at least have a minimal? */
if (adapter->num_tx_desc_avail <= EM_TX_OP_THRESHOLD) {
adapter->no_tx_desc_avail1++;
return (ENOBUFS);
}
}
/*
* Map the packet for DMA.
*
* 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.
*/
first = adapter->next_avail_tx_desc;
tx_buffer = &adapter->tx_buffer_area[first];
tx_buffer_mapped = tx_buffer;
map = tx_buffer->map;
error = bus_dmamap_load_mbuf_sg(adapter->txtag, map,
*m_headp, segs, &nsegs, BUS_DMA_NOWAIT);
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;
/* Try it again */
error = bus_dmamap_load_mbuf_sg(adapter->txtag, map,
*m_headp, segs, &nsegs, BUS_DMA_NOWAIT);
if (error == ENOMEM) {
adapter->no_tx_dma_setup++;
return (error);
} else if (error != 0) {
adapter->no_tx_dma_setup++;
m_freem(*m_headp);
*m_headp = NULL;
return (error);
}
} else if (error == ENOMEM) {
adapter->no_tx_dma_setup++;
return (error);
} else if (error != 0) {
adapter->no_tx_dma_setup++;
m_freem(*m_headp);
*m_headp = NULL;
return (error);
}
/* Check again to be sure we have enough descriptors */
if (nsegs > (adapter->num_tx_desc_avail - 2)) {
adapter->no_tx_desc_avail2++;
bus_dmamap_unload(adapter->txtag, map);
return (ENOBUFS);
}
m_head = *m_headp;
/*
* Set up the context descriptor:
* used when any hardware offload is done.
* This includes CSUM, VLAN, and TSO. It
* will use the first descriptor.
*/
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
/* First try TSO */
if (em_tso_adv_setup(adapter, m_head, &paylen)) {
cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
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
olinfo_status |= paylen << E1000_ADVTXD_PAYLEN_SHIFT;
} else if (m_head->m_pkthdr.csum_flags & CSUM_OFFLOAD) {
if (em_tx_adv_ctx_setup(adapter, m_head))
olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 */
i = adapter->next_avail_tx_desc;
for (j = 0; j < nsegs; j++) {
bus_size_t seg_len;
bus_addr_t seg_addr;
tx_buffer = &adapter->tx_buffer_area[i];
txd = (union e1000_adv_tx_desc *)&adapter->tx_desc_base[i];
seg_addr = segs[j].ds_addr;
seg_len = segs[j].ds_len;
txd->read.buffer_addr = htole64(seg_addr);
txd->read.cmd_type_len = htole32(
adapter->txd_cmd | cmd_type_len | seg_len);
txd->read.olinfo_status = htole32(olinfo_status);
last = i;
if (++i == adapter->num_tx_desc)
i = 0;
tx_buffer->m_head = NULL;
tx_buffer->next_eop = -1;
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
}
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->next_avail_tx_desc = i;
adapter->num_tx_desc_avail -= nsegs;
tx_buffer->m_head = m_head;
tx_buffer_mapped->map = tx_buffer->map;
tx_buffer->map = map;
bus_dmamap_sync(adapter->txtag, map, BUS_DMASYNC_PREWRITE);
/*
* Last Descriptor of Packet
* needs End Of Packet (EOP)
* and Report Status (RS)
*/
txd->read.cmd_type_len |=
htole32(E1000_TXD_CMD_EOP | E1000_TXD_CMD_RS);
/*
* Keep track in the first buffer which
* descriptor will be written back
*/
tx_buffer = &adapter->tx_buffer_area[first];
tx_buffer->next_eop = last;
/*
* Advance the Transmit Descriptor Tail (TDT), this tells the E1000
* that this frame is available to transmit.
*/
bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
E1000_WRITE_REG(&adapter->hw, E1000_TDT, 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
return (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
- 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
}
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;
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
uint16_t hw_tdt;
uint16_t sw_tdt;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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;
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
uint16_t length = 0;
boolean_t eop = 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
EM_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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_tdt = E1000_READ_REG(&adapter->hw, E1000_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
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
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_TDT, 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
{
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
uint32_t tctl;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_TDT) ==
E1000_READ_REG(&adapter->hw, E1000_TDH)) &&
(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;
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
uint32_t 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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
{
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
uint32_t 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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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;
uint32_t reg_rctl = 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
uint8_t 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_mc_addr_list_update(&adapter->hw, mta,
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
- 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_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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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);
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);
em_smartspeed(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
/*
* 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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
}
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
{
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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;
Call if_link_state_change() when link status changes. PR: kern/76890 Reviewed by: rwatson, sam
2005-02-04 18:36:04 +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
if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
E1000_STATUS_LU) {
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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 == 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_get_speed_and_duplex(&adapter->hw,
&adapter->link_speed, &adapter->link_duplex);
/* Check if we must disable SPEED_MODE bit on PCI-E */
if ((adapter->link_speed != SPEED_1000) &&
((adapter->hw.mac.type == e1000_82571) ||
(adapter->hw.mac.type == e1000_82572))) {
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
int tarc0;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
tarc0 = E1000_READ_REG(&adapter->hw,
E1000_TARC0);
tarc0 &= ~SPEED_MODE_BIT;
E1000_WRITE_REG(&adapter->hw,
E1000_TARC0, tarc0);
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
}
Put the "Link is up/down" printfs behind bootverbose. gigE is not so uncommon that we need to tell people about every cable in the network anymore. It can be enabled for debugging purposes with "boot -v".
2004-11-03 14:11:18 +00:00
if (bootverbose)
- Second style(9) 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, "Link is up %d Mbps %s\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->link_speed,
((adapter->link_duplex == FULL_DUPLEX) ?
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
"Full Duplex" : "Half Duplex"));
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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->link_active = 1;
adapter->smartspeed = 0;
ifp->if_baudrate = adapter->link_speed * 1000000;
Call if_link_state_change() when link status changes. PR: kern/76890 Reviewed by: rwatson, sam
2005-02-04 18:36:04 +00:00
if_link_state_change(ifp, LINK_STATE_UP);
- 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
if (adapter->link_active == 1) {
ifp->if_baudrate = adapter->link_speed = 0;
adapter->link_duplex = 0;
Put the "Link is up/down" printfs behind bootverbose. gigE is not so uncommon that we need to tell people about every cable in the network anymore. It can be enabled for debugging purposes with "boot -v".
2004-11-03 14:11:18 +00:00
if (bootverbose)
- Second style(9) 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, "Link is Down\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->link_active = 0;
Call if_link_state_change() when link status changes. PR: kern/76890 Reviewed by: rwatson, sam
2005-02-04 18:36:04 +00:00
if_link_state_change(ifp, LINK_STATE_DOWN);
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
}
}
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
}
/*********************************************************************
*
- Set RS (Report Status) bit on all descriptors of a packet instead of just the last one. - Set RDTR to zero by default instead of 28. - Fixed a problem with TX hangs with jumbo frames when number of fragments in the mbuf chain is large. - Added support for 82540EP based cards. MFC after: 3 days
2002-11-08 18:14:17 +00:00
* This routine disables all traffic on the adapter by issuing a
An attemp to make driver more readable and attaractive for further hacking: - Remove all spaces at eol. - Improve style(9) in most frequently edited functions. - In em_encap() push variables for 82544 workaround in the block where they are only used. - In em_get_buf() remove unused variable.
2006-01-20 11:38:25 +00:00
* global reset on the MAC and deallocates TX/RX buffers.
This is the first commit of the Intel gigabit driver for PRO/1000 cards. Submitted by:Prafulla Deuskar Reviewed by: Paul Saab MFC after:1 week
2001-12-02 07:37:17 +00:00
*
**********************************************************************/
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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_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
- 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
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);
if ((adapter->hw.bus.pci_cmd_word & PCIM_CMD_BUSMASTEREN) == 0 &&
(adapter->hw.bus.pci_cmd_word & PCIM_CMD_MEMEN)) {
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);
adapter->hw.subsystem_device_id = pci_read_config(dev, PCIR_SUBDEV_0, 2);
/* 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;
- Second style(9) 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 val, rid;
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
Make em(4) work on big-endian architectures. - disable jumbo frame support on strict alignment architectures due to the limitation of hardware. The driver needs a fix-up code for RX side. The fix will show up in near future. - fix endian issue for 82544 on PCI-X bus. I couldn't test this as I don't have the NIC/hardware. - prefer PCIR_BAR to hardcoded EM_MMBA. - Properly checks for for 64bit BAR [1] - replace inl/outl with bus_space(9) [1] - fix endian issue on VLAN handling. - reorder header files and remove unnecessary one. Reviewed by: cognet No response from: pdeuskar, tackerman Obtained from: OpenBSD [1]
2005-11-09 08:43:18 +00:00
rid = PCIR_BAR(0);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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->res_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);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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->res_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 =
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
rman_get_bustag(adapter->res_memory);
adapter->osdep.mem_bus_space_handle =
rman_get_bushandle(adapter->res_memory);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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.hw_addr = (uint8_t *)&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 */
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
if ((adapter->hw.mac.type > e1000_82542) &&
(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
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->res_ioport = bus_alloc_resource_any(dev,
SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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->res_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 =
rman_get_bustag(adapter->res_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 =
rman_get_bushandle(adapter->res_ioport);
- 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
* only the latest can use MSI/X and
* real support for it is forthcoming
*/
adapter->msi = 0; /* Set defaults */
rid = 0x0;
if (adapter->hw.mac.type >= e1000_82575) {
/*
* Eventually this will be used
* for Multiqueue, for now we will
* just use one vector.
*/
val = pci_msix_count(dev);
if ((val) && pci_alloc_msix(dev, &val) == 0) {
rid = 1;
adapter->msi = 1;
}
} else if (adapter->hw.bus.type == e1000_bus_type_pci_express) {
val = pci_msi_count(dev);
if (val == 1 && pci_alloc_msi(dev, &val) == 0) {
rid = 1;
adapter->msi = 1;
}
}
adapter->res_interrupt = bus_alloc_resource_any(dev,
SYS_RES_IRQ, &rid, RF_SHAREABLE | RF_ACTIVE);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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->res_interrupt == 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: "
"interrupt\n");
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
}
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
- Second style(9) 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);
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
/*********************************************************************
*
* Setup the appropriate Interrupt handlers.
*
**********************************************************************/
int
em_allocate_intr(struct adapter *adapter)
{
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
#ifdef DEVICE_POLLING
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 */
if (adapter->int_handler_tag == NULL &&
(error = bus_setup_intr(dev, adapter->res_interrupt,
INTR_TYPE_NET | INTR_MPSAFE, NULL, em_intr, adapter,
&adapter->int_handler_tag)) != 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
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
#else
/*
* 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));
if ((error = bus_setup_intr(dev, adapter->res_interrupt,
Revert previous change and take back a pointy hat.
2007-02-28 09:04:46 +00:00
INTR_TYPE_NET, em_intr_fast, NULL, adapter,
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
&adapter->int_handler_tag)) != 0) {
device_printf(dev, "Failed to register fast interrupt "
"handler: %d\n", error);
taskqueue_free(adapter->tq);
adapter->tq = NULL;
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
#endif
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_enable_intr(adapter);
return (0);
}
Fix the interrupt race for real. Don't register the interrupt until after the the interface has been configured. I'm not sure how this could ever have worked before, but it should be fixed now. Also break out the interrupt degresitration function into it's own step.
2006-01-13 08:18:04 +00:00
static void
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_free_intr(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;
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
if (adapter->res_interrupt != NULL) {
bus_teardown_intr(dev, adapter->res_interrupt,
adapter->int_handler_tag);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
adapter->int_handler_tag = NULL;
This is the merge of the Intel 6.2.9 driver. It provides all new shared code, new device support, and it is hoped a more stable driver for 6.2. RELEASE. This checkin was discussed and approved today by RE, scottl, jhb, and pdeuskar
2006-10-28 00:47:55 +00:00
}
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
if (adapter->tq != NULL) {
taskqueue_drain(adapter->tq, &adapter->rxtx_task);
taskqueue_drain(taskqueue_fast, &adapter->link_task);
taskqueue_free(adapter->tq);
adapter->tq = NULL;
}
}
static void
em_free_pci_resources(struct adapter *adapter)
{
device_t dev = adapter->dev;
if (adapter->res_interrupt != 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_IRQ,
0, adapter->res_interrupt);
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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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->res_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,
PCIR_BAR(0), adapter->res_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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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->flash_mem != 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,
EM_FLASH, adapter->flash_mem);
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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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->res_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,
adapter->io_rid, adapter->res_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
}
/*********************************************************************
*
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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;
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
uint16_t rx_buffer_size;
- Second style(9) 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
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)) {
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
uint16_t phy_tmp = 0;
/* 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 );
adapter->hw.mac.fc_high_water = rx_buffer_size -
roundup2(adapter->hw.mac.max_frame_size, 1024);
adapter->hw.mac.fc_low_water = adapter->hw.mac.fc_high_water - 1500;
if (adapter->hw.mac.type == e1000_80003es2lan)
adapter->hw.mac.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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.fc_pause_time = EM_FC_PAUSE_TIME;
adapter->hw.mac.fc_send_xon = TRUE;
adapter->hw.mac.fc = 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) {
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
ifp->if_capabilities |= IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM;
ifp->if_capenable |= IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM;
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
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
/* Enable TSO for PCI Express adapters */
if (adapter->hw.bus.type == e1000_bus_type_pci_express) {
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
ifp->if_capabilities |= IFCAP_TSO4;
ifp->if_capenable |= IFCAP_TSO4;
Fix style(9) issues in the TSO specific changes. Pointed out by: jmallett
2006-09-10 19:23:27 +00:00
}
Add support for TSO. Thanks to Andre for adding support in the stack and Jack Vogel for driver changes. Submitted by: Jack Vogel
2006-09-09 06:19:20 +00:00
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_capenable |= 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);
if ((adapter->hw.media_type == e1000_media_type_fiber) ||
(adapter->hw.media_type == e1000_media_type_internal_serdes)) {
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
{
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
uint16_t 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) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
if(phy_tmp & CR_1000T_MS_ENABLE) {
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,
- Second style(9) 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);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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) &&
!e1000_read_phy_reg(&adapter->hw, PHY_CONTROL,
- Second style(9) 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)) {
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,
- Second style(9) 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);
}
}
}
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;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(adapter->dev), /* parent */
Merge in new driver from Intel, version 6.0.5. It adds support for 80003 NICs and NICs found on ICH8 mobos, and improves support for already known chips. Details: - if_em.c. Merged manually, viewing diff between new vendor driver and previous one. This was an easy task, because most changes between 5.1.5 and 6.0.5 are bugfixes taken from FreeBSD. - if_em_hw.h. Dropped in from vendor, and then restored revisions 1.16, 1.17, 1.18. - if_em_hw.c. Dropped in from vendor, and then restored revision 1.15. - if_em_osdep.h. Added new required macros from vendor file and add a hack against define namespace mangling in if_em_hw.h. Intel made another hack, but I prefer mine.
2006-08-03 09:20:11 +00:00
EM_DBA_ALIGN, 0, /* alignment, bounds */
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
size, /* maxsize */
1, /* nsegments */
size, /* maxsegsize */
0, /* flags */
NULL, /* lockfunc */
NULL, /* lockarg */
&dma->dma_tag);
if (error) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
device_printf(adapter->dev,
"%s: bus_dma_tag_create failed: %d\n",
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
__func__, error);
goto fail_0;
}
error = bus_dmamem_alloc(dma->dma_tag, (void**) &dma->dma_vaddr,
BUS_DMA_NOWAIT, &dma->dma_map);
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)
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_t dev = adapter->dev;
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");
return (ENOMEM);
}
bzero(adapter->tx_buffer_area,
(sizeof(struct em_buffer)) * adapter->num_tx_desc);
return (0);
}
/*********************************************************************
*
* Initialize transmit structures.
*
**********************************************************************/
static int
em_setup_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;
int error, 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
/*
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
*/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 */
- Second style(9) 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 = em_allocate_transmit_structures(adapter)) != 0)
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;
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
/* Clear the old ring contents */
bzero(adapter->tx_desc_base,
(sizeof(struct e1000_tx_desc)) * adapter->num_tx_desc);
/* Create the descriptor buffer dma maps */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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;
for (i = 0; i < adapter->num_tx_desc; i++) {
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
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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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;
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
/* Set number of descriptors available */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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
return (0);
fail:
em_free_transmit_structures(adapter);
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
}
/*********************************************************************
*
* 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
{
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
uint32_t tctl, tarc, tipg = 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
uint64_t 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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_TDLEN,
adapter->num_tx_desc * sizeof(struct e1000_tx_desc));
E1000_WRITE_REG(&adapter->hw, E1000_TDBAH, (uint32_t)(bus_addr >> 32));
E1000_WRITE_REG(&adapter->hw, E1000_TDBAL, (uint32_t)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 */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_TDT, 0);
E1000_WRITE_REG(&adapter->hw, E1000_TDH, 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",
E1000_READ_REG(&adapter->hw, E1000_TDBAL),
E1000_READ_REG(&adapter->hw, E1000_TDLEN));
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:
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.media_type == e1000_media_type_fiber) ||
(adapter->hw.media_type ==
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)) {
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC0);
tarc |= SPEED_MODE_BIT;
E1000_WRITE_REG(&adapter->hw, E1000_TARC0, tarc);
} else if (adapter->hw.mac.type == e1000_80003es2lan) {
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC0);
tarc |= 1;
E1000_WRITE_REG(&adapter->hw, E1000_TARC0, tarc);
tarc = E1000_READ_REG(&adapter->hw, E1000_TARC1);
tarc |= 1;
E1000_WRITE_REG(&adapter->hw, E1000_TARC1, tarc);
}
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;
tctl = E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
(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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->tx_int_delay.value > 0) &&
(adapter->hw.mac.type != e1000_82575))
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 the function pointer for the transmit routine */
if (adapter->hw.mac.type >= e1000_82575)
adapter->em_xmit = em_adv_encap;
else
adapter->em_xmit = em_encap;
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;
int i;
- 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) {
tx_buffer = adapter->tx_buffer_area;
for (i = 0; i < adapter->num_tx_desc; i++, tx_buffer++) {
- Second style(9) 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,
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
uint32_t *txd_upper, uint32_t *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;
struct ip *ip;
struct ip6_hdr *ip6;
struct tcp_hdr *th;
int curr_txd, ehdrlen, hdr_len, ip_hlen;
uint32_t cmd = 0;
uint16_t etype;
uint8_t ipproto;
/* 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
*txd_lower = E1000_TXD_CMD_DEXT | /* Extended descr type */
E1000_TXD_DTYP_D; /* Data descr */
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;
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:
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;
}
break;
default:
break;
}
This is the merge of the Intel 6.2.9 driver. It provides all new shared code, new device support, and it is hoped a more stable driver for 6.2. RELEASE. This checkin was discussed and approved today by RE, scottl, jhb, and pdeuskar
2006-10-28 00:47:55 +00:00
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
TXD->tcp_seg_setup.data = htole32(0);
TXD->cmd_and_length =
htole32(adapter->txd_cmd | E1000_TXD_CMD_DEXT | cmd);
- 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
}
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)
*
**********************************************************************/
static boolean_t
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
em_tso_setup(struct adapter *adapter, struct mbuf *mp, uint32_t *txd_upper,
uint32_t *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;
uint16_t etype;
/*
* XXX: This is not really correct as the stack would not have
* set up all checksums.
* XXX: Return FALSE is not sufficient as we may have to return
* in true failure cases as well. Should do -1 (failure), 0 (no)
* and 1 (success).
*/
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 (((mp->m_pkthdr.csum_flags & CSUM_TSO) == 0) ||
(mp->m_pkthdr.len <= EM_TX_BUFFER_SIZE))
return FALSE;
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
/*
* 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
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 work for hardware segmentation offload (TSO) on
* adapters using advanced tx descriptors
*
**********************************************************************/
static boolean_t
em_tso_adv_setup(struct adapter *adapter, struct mbuf *mp, u32 *paylen)
{
struct e1000_adv_tx_context_desc *TXD;
struct em_buffer *tx_buffer;
u32 vlan_macip_lens = 0, type_tucmd_mlhl = 0;
u32 mss_l4len_idx = 0;
u16 vtag = 0;
int ctxd, ehdrlen, hdrlen, ip_hlen, tcp_hlen;
struct ether_vlan_header *eh;
struct ip *ip;
struct tcphdr *th;
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
if (((mp->m_pkthdr.csum_flags & CSUM_TSO) == 0) ||
(mp->m_pkthdr.len <= EM_TX_BUFFER_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
return FALSE;
/*
* Determine where frame payload starts.
* Jump over vlan headers if already present
*/
eh = mtod(mp, struct ether_vlan_header *);
if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN))
ehdrlen = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
else
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;
/* Only supports IPV4 for now */
ctxd = adapter->next_avail_tx_desc;
tx_buffer = &adapter->tx_buffer_area[ctxd];
TXD = (struct e1000_adv_tx_context_desc *) &adapter->tx_desc_base[ctxd];
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;
th = (struct tcphdr *)((caddr_t)ip + ip_hlen);
th->th_sum = in_pseudo(ip->ip_src.s_addr,
ip->ip_dst.s_addr, htons(IPPROTO_TCP));
tcp_hlen = th->th_off << 2;
hdrlen = ehdrlen + ip_hlen + tcp_hlen;
/* Calculate payload, this is used in the transmit desc in encap */
*paylen = mp->m_pkthdr.len - hdrlen;
/* VLAN MACLEN IPLEN */
if (mp->m_flags & M_VLANTAG) {
vtag = htole16(mp->m_pkthdr.ether_vtag);
vlan_macip_lens |= (vtag << E1000_ADVTXD_VLAN_SHIFT);
}
vlan_macip_lens |= (ehdrlen << E1000_ADVTXD_MACLEN_SHIFT);
vlan_macip_lens |= ip_hlen;
TXD->vlan_macip_lens |= htole32(vlan_macip_lens);
/* ADV DTYPE TUCMD */
type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4;
TXD->type_tucmd_mlhl |= htole32(type_tucmd_mlhl);
/* MSS L4LEN IDX */
mss_l4len_idx |= (mp->m_pkthdr.tso_segsz << E1000_ADVTXD_MSS_SHIFT);
mss_l4len_idx |= (tcp_hlen << E1000_ADVTXD_L4LEN_SHIFT);
TXD->mss_l4len_idx = htole32(mss_l4len_idx);
TXD->seqnum_seed = htole32(0);
tx_buffer->m_head = NULL;
tx_buffer->next_eop = -1;
if (++ctxd == adapter->num_tx_desc)
ctxd = 0;
adapter->num_tx_desc_avail--;
adapter->next_avail_tx_desc = ctxd;
return TRUE;
}
/*********************************************************************
*
* Advanced Context Descriptor setup for VLAN or CSUM
*
**********************************************************************/
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
static boolean_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
em_tx_adv_ctx_setup(struct adapter *adapter, struct mbuf *mp)
{
struct e1000_adv_tx_context_desc *TXD;
struct em_buffer *tx_buffer;
uint32_t vlan_macip_lens = 0, type_tucmd_mlhl = 0;
struct ether_vlan_header *eh;
struct ip *ip;
struct ip6_hdr *ip6;
int ehdrlen, ip_hlen;
u16 etype;
u8 ipproto;
int ctxd = adapter->next_avail_tx_desc;
u16 vtag = 0;
tx_buffer = &adapter->tx_buffer_area[ctxd];
TXD = (struct e1000_adv_tx_context_desc *) &adapter->tx_desc_base[ctxd];
/*
** In advanced descriptors the vlan tag must
** be placed into the descriptor itself.
*/
if (mp->m_flags & M_VLANTAG) {
vtag = htole16(mp->m_pkthdr.ether_vtag);
vlan_macip_lens |= (vtag << E1000_ADVTXD_VLAN_SHIFT);
}
/*
* 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;
}
/* Set the ether header length */
vlan_macip_lens |= ehdrlen << E1000_ADVTXD_MACLEN_SHIFT;
switch (etype) {
case ETHERTYPE_IP:
ip = (struct ip *)(mp->m_data + ehdrlen);
ip_hlen = ip->ip_hl << 2;
if (mp->m_len < ehdrlen + ip_hlen)
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
return FALSE; /* 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
ipproto = ip->ip_p;
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4;
break;
case ETHERTYPE_IPV6:
ip6 = (struct ip6_hdr *)(mp->m_data + ehdrlen);
ip_hlen = sizeof(struct ip6_hdr);
if (mp->m_len < ehdrlen + ip_hlen)
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
return FALSE; /* 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
ipproto = ip6->ip6_nxt;
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV6;
break;
default:
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
return 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
}
vlan_macip_lens |= ip_hlen;
type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
switch (ipproto) {
case IPPROTO_TCP:
if (mp->m_pkthdr.csum_flags & CSUM_TCP)
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
break;
case IPPROTO_UDP:
if (mp->m_pkthdr.csum_flags & CSUM_UDP)
type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
break;
}
/* Now copy bits into descriptor */
TXD->vlan_macip_lens |= htole32(vlan_macip_lens);
TXD->type_tucmd_mlhl |= htole32(type_tucmd_mlhl);
TXD->seqnum_seed = htole32(0);
TXD->mss_l4len_idx = htole32(0);
tx_buffer->m_head = NULL;
tx_buffer->next_eop = -1;
/* We've consumed the first desc, adjust counters */
if (++ctxd == adapter->num_tx_desc)
ctxd = 0;
adapter->next_avail_tx_desc = ctxd;
--adapter->num_tx_desc_avail;
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
return TRUE;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
}
- 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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_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 */
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 */
else 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;
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
}
/*********************************************************************
*
* 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
if (adapter->hw.mac.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) *
adapter->num_rx_desc, M_DEVBUF, M_NOWAIT);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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);
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_buffer_area,
sizeof(struct em_buffer) * adapter->num_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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_dma_tag_create(bus_get_dma_tag(dev), /* parent */
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
1, 0, /* alignment, bounds */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
MCLBYTES, /* maxsize */
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
1, /* nsegments */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
MCLBYTES, /* maxsegsize */
Nuke invalid use of BUS_DMA_ALLOCNOW.
2006-07-27 00:29:44 +00:00
0, /* flags */
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
NULL, /* lockfunc */
NULL, /* lockarg */
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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;
}
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 initial buffers */
for (i = 0; i < adapter->num_rx_desc; i++) {
error = em_get_buf(adapter, i);
if (error)
goto fail;
}
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);
- 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
}
/*********************************************************************
*
* Allocate and 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
{
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 = em_allocate_receive_structures(adapter)) !=0)
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;
- Second style(9) 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
**********************************************************************/
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;
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
uint64_t bus_addr;
uint32_t reg_rctl;
uint32_t reg_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
*/
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl & ~E1000_RCTL_EN);
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_82540) {
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)
*/
#define MAX_INTS_PER_SEC 8000
#define DEFAULT_ITR 1000000000/(MAX_INTS_PER_SEC * 256)
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_RDLEN, adapter->num_rx_desc *
sizeof(struct e1000_rx_desc));
E1000_WRITE_REG(&adapter->hw, E1000_RDBAH, (uint32_t)(bus_addr >> 32));
E1000_WRITE_REG(&adapter->hw, E1000_RDBAL, (uint32_t)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 */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 &= ~(3 << E1000_RCTL_MO_SHIFT);
reg_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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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))
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +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
else
reg_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:
- 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
reg_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:
reg_rctl |= E1000_RCTL_SZ_4096 |
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:
reg_rctl |= E1000_RCTL_SZ_8192 |
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:
reg_rctl |= E1000_RCTL_SZ_16384 |
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)
reg_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
reg_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)) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_rxcsum = E1000_READ_REG(&adapter->hw, E1000_RXCSUM);
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
reg_rxcsum |= (E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_RXCSUM, reg_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 */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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);
/*
* Setup the HW Rx Head and
* Tail Descriptor Pointers
*/
E1000_WRITE_REG(&adapter->hw, E1000_RDH, 0);
E1000_WRITE_REG(&adapter->hw, E1000_RDT, 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
{
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 ifnet *ifp;
struct mbuf *mp;
uint8_t accept_frame = 0;
uint8_t eop = 0;
uint16_t len, desc_len, prev_len_adj;
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
/* Pointer to the receive descriptor being examined. */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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;
Don't update Rx descriptor status in two different functions. Suggested by: pdeuskar Reviewed by: pdeuskar
2006-08-16 23:55:34 +00:00
uint8_t status;
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
ifp = adapter->ifp;
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
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);
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) {
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
uint8_t last_byte;
uint32_t 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,
current_desc->errors, pkt_len, last_byte)) {
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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,
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.addr);
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.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) {
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;
}
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.max_frame_size <=
(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;
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
#ifdef 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_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);
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +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
/* Already running unlocked */
Backout bogus checkin to HEAD Approved by: scottl
2006-10-28 08:11:07 +00:00
(*ifp->if_input)(ifp, m);
#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
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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_RDT, 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
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
}
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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_vlans(struct adapter *adapter)
Added support for 82545EM and 82546EB based adapters. Added Vlan support. MFC after: 1 week
2002-06-03 22:30:51 +00:00
{
uint32_t ctrl;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_VET, ETHERTYPE_VLAN);
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);
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
{
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_IMS,
(IMS_ENABLE_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
{
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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:
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:
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
- Second style(9) 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_is_valid_ether_addr(uint8_t *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
/*
* NOTE: the following routines using the e1000
* naming style are provided to the shared
* code which expects that rather than 'em'
*/
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
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
e1000_write_pci_cfg(struct e1000_hw *hw, uint32_t reg, uint16_t *value)
- 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
{
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
pci_write_config(((struct e1000_osdep *)hw->back)->dev, reg, *value, 2);
- 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
}
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
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
e1000_read_pci_cfg(struct e1000_hw *hw, uint32_t reg, uint16_t *value)
- 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
*value = pci_read_config(((struct e1000_osdep *)hw->back)->dev, reg, 2);
- 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
}
- 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
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
e1000_pci_set_mwi(struct e1000_hw *hw)
- 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
pci_write_config(((struct e1000_osdep *)hw->back)->dev, PCIR_COMMAND,
(hw->bus.pci_cmd_word | CMD_MEM_WRT_INVALIDATE), 2);
- 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
}
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
e1000_pci_clear_mwi(struct e1000_hw *hw)
- 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
pci_write_config(((struct e1000_osdep *)hw->back)->dev, PCIR_COMMAND,
(hw->bus.pci_cmd_word & ~CMD_MEM_WRT_INVALIDATE), 2);
- 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
/*
* Read the PCI Express capabilities
*/
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
int32_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
e1000_read_pcie_cap_reg(struct e1000_hw *hw, uint32_t reg, uint16_t *value)
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
int32_t error = E1000_SUCCESS;
uint16_t cap_off;
switch (hw->mac.type) {
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_82573:
case e1000_80003es2lan:
cap_off = 0xE0;
e1000_read_pci_cfg(hw, cap_off + reg, value);
break;
default:
error = ~E1000_NOT_IMPLEMENTED;
break;
}
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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 (error);
}
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
int32_t
e1000_alloc_zeroed_dev_spec_struct(struct e1000_hw *hw, uint32_t size)
{
int32_t error = 0;
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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->dev_spec = malloc(size, M_DEVBUF, M_NOWAIT | M_ZERO);
if (hw->dev_spec == NULL)
error = ENOMEM;
bzero(hw->dev_spec, size);
return (error);
}
void
e1000_free_dev_spec_struct(struct e1000_hw *hw)
{
if (hw->dev_spec != NULL)
free(hw->dev_spec, M_DEVBUF);
return;
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
/*
* 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
*************************************************************************/
- Second style(9) 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 uint32_t
em_fill_descriptors (bus_addr_t address, uint32_t length,
PDESC_ARRAY desc_array)
{
/* Since issue is sensitive to length and address.*/
/* Let us first check the address...*/
uint32_t safe_terminator;
if (length <= 4) {
desc_array->descriptor[0].address = address;
desc_array->descriptor[0].length = length;
desc_array->elements = 1;
return (desc_array->elements);
}
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance if any thing gets clobbered, I'm sure I'll hear about it :) Approved by pdeuskar
2007-05-04 00:00:12 +00:00
safe_terminator = (uint32_t)((((uint32_t)address & 0x7) +
(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;
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.media_type == e1000_media_type_copper ||
(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 */
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.gorcl += E1000_READ_REG(&adapter->hw, E1000_GORCL);
adapter->stats.gorch += E1000_READ_REG(&adapter->hw, E1000_GORCH);
adapter->stats.gotcl += E1000_READ_REG(&adapter->hw, E1000_GOTCL);
adapter->stats.gotch += E1000_READ_REG(&adapter->hw, E1000_GOTCH);
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);
adapter->stats.torl += E1000_READ_REG(&adapter->hw, E1000_TORL);
adapter->stats.torh += E1000_READ_REG(&adapter->hw, E1000_TORH);
adapter->stats.totl += E1000_READ_REG(&adapter->hw, E1000_TOTL);
adapter->stats.toth += E1000_READ_REG(&adapter->hw, E1000_TOTH);
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;
uint8_t *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",
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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.fc_high_water,
adapter->hw.mac.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",
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_TDH),
E1000_READ_REG(&adapter->hw, E1000_TDT));
device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
E1000_READ_REG(&adapter->hw, E1000_RDH),
E1000_READ_REG(&adapter->hw, E1000_RDT));
- Second style(9) 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);
- Second style(9) 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, "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);
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
}
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
}
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;
- Second style(9) megacleanup. - Rename "adapter" to "sc"/"softc", to be like other drivers. (-13 Kb less source code)
2006-02-15 08:39:50 +00:00
uint32_t 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
Revert back changes to made in rev 1.109 of if_em.c which were unnecessary. 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_LOCK(adapter);
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
Merge in the new driver (6.5.0) of Intel. This has a new shared code infrastructure that is family specific and modular. There is also support for our latest gigabit nic, the 82575 that is MSI/X and multiqueue capable. The new shared code changes some interfaces to the core code but testing at Intel has been going on for months, it is fairly stable. I have attempted to be careful in retaining any fixes that CURRENT had and we did not, I apologize in advance 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_82575)
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);
EM_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
#ifndef DEVICE_POLLING
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
Reference in New Issue Copy Permalink