d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
9d307c8df9
freebsd-skq/sys/dev/fxp/if_fxp.c

3168 lines
90 KiB
C
Raw Normal View History

Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/*-
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
* Copyright (c) 1995, David Greenman
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
* Copyright (c) 2001 Jonathan Lemon <jlemon@freebsd.org>
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice unmodified, this list of conditions, and the following
* disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*/
Use __FBSDID(). Also some minor style cleanups.
2003-08-24 17:55:58 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
This driver really is for the Pro/100B and not for the Pro/100 (and in fact doesn't even work with a Pro/100)...updated to reflect this.
1996-01-15 10:12:41 +00:00
* Intel EtherExpress Pro/100B PCI Fast Ethernet driver
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +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
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
#include <sys/param.h>
#include <sys/systm.h>
Sort head files and removed ununsed header file.
2008-11-25 00:59:11 +00:00
#include <sys/bus.h>
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
#include <sys/endian.h>
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
#include <sys/kernel.h>
Sort head files and removed ununsed header file.
2008-11-25 00:59:11 +00:00
#include <sys/mbuf.h>
mutex.h is needed here. It got it by namespace pollution. Pointed out by: bde
2008-12-08 03:48:03 +00:00
#include <sys/lock.h>
Add missing <sys/module.h> includes
2004-05-30 20:08:47 +00:00
#include <sys/module.h>
mutex.h is needed here. It got it by namespace pollution. Pointed out by: bde
2008-12-08 03:48:03 +00:00
#include <sys/mutex.h>
Sort head files and removed ununsed header file.
2008-11-25 00:59:11 +00:00
#include <sys/rman.h>
Removed nested include if <sys/socket.h> from <net/if.h> and <net/if_arp.h> and fixed the things that depended on it. The nested include just allowed unportable programs to compile and made my simple #include checking program report that networking code doesn't need to include <sys/socket.h>.
1996-10-12 19:49:43 +00:00
#include <sys/socket.h>
Sort head files and removed ununsed header file.
2008-11-25 00:59:11 +00:00
#include <sys/sockio.h>
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
#include <sys/sysctl.h>
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Sort head files and removed ununsed header file.
2008-11-25 00:59:11 +00:00
#include <net/bpf.h>
#include <net/ethernet.h>
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
#include <net/if.h>
Sort head files and removed ununsed header file.
2008-11-25 00:59:11 +00:00
#include <net/if_arp.h>
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
#include <net/if_dl.h>
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
#include <net/if_media.h>
Sort head files and removed ununsed header file.
2008-11-25 00:59:11 +00:00
#include <net/if_types.h>
#include <net/if_vlan_var.h>
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
#include <netinet/tcp.h>
#include <netinet/udp.h>
#include <machine/bus.h>
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
#include <machine/in_cksum.h>
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
#include <machine/resource.h>
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
Prefer new location of pci include files (which have only been in the tree for two or more years now), except in a few places where there's code to be compatible with older versions of FreeBSD.
2003-08-22 05:54:52 +00:00
#include <dev/pci/pcivar.h>
#include <dev/pci/pcireg.h> /* for PCIM_CMD_xxx */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
#include <dev/mii/mii.h>
#include <dev/mii/miivar.h>
#include <dev/fxp/if_fxpreg.h>
#include <dev/fxp/if_fxpvar.h>
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
#include <dev/fxp/rcvbundl.h>
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
- Express hard dependencies on bus (pci, isa, pccard) and network layer (ether). - Don't abuse module names to facilitate ifconfig module loading; such abuse isn't really needed. (And if we do need type information associated with a module then we should make it explicit and not use hacks.)
2003-04-15 06:37:30 +00:00
MODULE_DEPEND(fxp, pci, 1, 1, 1);
MODULE_DEPEND(fxp, ether, 1, 1, 1);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
MODULE_DEPEND(fxp, miibus, 1, 1, 1);
#include "miibus_if.h"
Make the fxp driver work on alpha, rather than panic the machine on boot and/or when using the card. o Convert the driver to using bus_space. This allows alphas with fxp's to boot, rather than panic'ing because rman_get_virtual() doesn't really return a virtual address on alphas. o Fix an alpha unaligned access error caused by some misfeature of gcc/egcs: if link_addr & rbd_addr in the fxp_rfa struct are 32 bit quantities, egcs will assume they are naturally aligned. So it will do a ldl & some shifty/masky to twiddle 16 bit values in fxp_lwcopy(). However, if they are 16-bit aligned, the ldl will actually be done on a 16-bit aligned value & we will panic with an unaligned access error... Changing their definition to an array of chars seems to fix this. I obtained this from NetBSD. I've tested this on both i386 & alpha.
1999-09-30 19:03:12 +00:00
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
/*
* NOTE! On the Alpha, we have an alignment constraint. The
* card DMAs the packet immediately following the RFA. However,
* the first thing in the packet is a 14-byte Ethernet header.
* This means that the packet is misaligned. To compensate,
* we actually offset the RFA 2 bytes into the cluster. This
* alignes the packet after the Ethernet header at a 32-bit
* boundary. HOWEVER! This means that the RFA is misaligned!
*/
#define RFA_ALIGNMENT_FUDGE 2
/*
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
* Set initial transmit threshold at 64 (512 bytes). This is
* increased by 64 (512 bytes) at a time, to maximum of 192
* (1536 bytes), if an underrun occurs.
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
*/
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static int tx_threshold = 64;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
* The configuration byte map has several undefined fields which
8255x configure command requires number of bytes of configuration table. The default size of the configuration table was 22 bytes. To use extended feature of 82550/82551 the configuration table size was expanded to 32 bytes. The added configuration for 82550/82551 specifies VLAN hardware tagging and IPSec configuration as well as TCO. To make configuration easier fxp(4) used a configuration template and the template was copied to configuration table. After that, some parameters of the configuration table was changed depending on controller type and operation mode. However the size of template was 22 bytes so some configuration parameters were not properly initialized on 82550/82551. Fix this by increasing the template size. For 82557, 82558 and 82559 the size of the configuration is still 22 bytes.
2010-05-07 18:03:35 +00:00
* must be one or must be zero. Set up a template for these bits.
* The actual configuration is performed in fxp_init.
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
*
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
* See struct fxp_cb_config for the bit definitions.
*/
static u_char fxp_cb_config_template[] = {
0x0, 0x0, /* cb_status */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
0x0, 0x0, /* cb_command */
0x0, 0x0, 0x0, 0x0, /* link_addr */
0x0, /* 0 */
0x0, /* 1 */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
0x0, /* 2 */
0x0, /* 3 */
0x0, /* 4 */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
0x0, /* 5 */
0x32, /* 6 */
0x0, /* 7 */
0x0, /* 8 */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
0x0, /* 9 */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
0x6, /* 10 */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
0x0, /* 11 */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
0x0, /* 12 */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
0x0, /* 13 */
0xf2, /* 14 */
0x48, /* 15 */
0x0, /* 16 */
0x40, /* 17 */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
0xf0, /* 18 */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
0x0, /* 19 */
0x3f, /* 20 */
8255x configure command requires number of bytes of configuration table. The default size of the configuration table was 22 bytes. To use extended feature of 82550/82551 the configuration table size was expanded to 32 bytes. The added configuration for 82550/82551 specifies VLAN hardware tagging and IPSec configuration as well as TCO. To make configuration easier fxp(4) used a configuration template and the template was copied to configuration table. After that, some parameters of the configuration table was changed depending on controller type and operation mode. However the size of template was 22 bytes so some configuration parameters were not properly initialized on 82550/82551. Fix this by increasing the template size. For 82557, 82558 and 82559 the size of the configuration is still 22 bytes.
2010-05-07 18:03:35 +00:00
0x5, /* 21 */
0x0, /* 22 */
0x0, /* 23 */
0x0, /* 24 */
0x0, /* 25 */
0x0, /* 26 */
0x0, /* 27 */
0x0, /* 28 */
0x0, /* 29 */
0x0, /* 30 */
0x0 /* 31 */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
};
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/*
* Claim various Intel PCI device identifiers for this driver. The
* sub-vendor and sub-device field are extensively used to identify
* particular variants, but we don't currently differentiate between
* them.
*/
static struct fxp_ident fxp_ident_table[] = {
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
{ 0x1029, -1, 0, "Intel 82559 PCI/CardBus Pro/100" },
{ 0x1030, -1, 0, "Intel 82559 Pro/100 Ethernet" },
{ 0x1031, -1, 3, "Intel 82801CAM (ICH3) Pro/100 VE Ethernet" },
{ 0x1032, -1, 3, "Intel 82801CAM (ICH3) Pro/100 VE Ethernet" },
{ 0x1033, -1, 3, "Intel 82801CAM (ICH3) Pro/100 VM Ethernet" },
{ 0x1034, -1, 3, "Intel 82801CAM (ICH3) Pro/100 VM Ethernet" },
{ 0x1035, -1, 3, "Intel 82801CAM (ICH3) Pro/100 Ethernet" },
{ 0x1036, -1, 3, "Intel 82801CAM (ICH3) Pro/100 Ethernet" },
{ 0x1037, -1, 3, "Intel 82801CAM (ICH3) Pro/100 Ethernet" },
{ 0x1038, -1, 3, "Intel 82801CAM (ICH3) Pro/100 VM Ethernet" },
{ 0x1039, -1, 4, "Intel 82801DB (ICH4) Pro/100 VE Ethernet" },
{ 0x103A, -1, 4, "Intel 82801DB (ICH4) Pro/100 Ethernet" },
{ 0x103B, -1, 4, "Intel 82801DB (ICH4) Pro/100 VM Ethernet" },
{ 0x103C, -1, 4, "Intel 82801DB (ICH4) Pro/100 Ethernet" },
{ 0x103D, -1, 4, "Intel 82801DB (ICH4) Pro/100 VE Ethernet" },
{ 0x103E, -1, 4, "Intel 82801DB (ICH4) Pro/100 VM Ethernet" },
{ 0x1050, -1, 5, "Intel 82801BA (D865) Pro/100 VE Ethernet" },
{ 0x1051, -1, 5, "Intel 82562ET (ICH5/ICH5R) Pro/100 VE Ethernet" },
{ 0x1059, -1, 0, "Intel 82551QM Pro/100 M Mobile Connection" },
{ 0x1064, -1, 6, "Intel 82562EZ (ICH6)" },
{ 0x1065, -1, 6, "Intel 82562ET/EZ/GT/GZ PRO/100 VE Ethernet" },
{ 0x1068, -1, 6, "Intel 82801FBM (ICH6-M) Pro/100 VE Ethernet" },
{ 0x1069, -1, 6, "Intel 82562EM/EX/GX Pro/100 Ethernet" },
{ 0x1091, -1, 7, "Intel 82562GX Pro/100 Ethernet" },
{ 0x1092, -1, 7, "Intel Pro/100 VE Network Connection" },
{ 0x1093, -1, 7, "Intel Pro/100 VM Network Connection" },
{ 0x1094, -1, 7, "Intel Pro/100 946GZ (ICH7) Network Connection" },
{ 0x1209, -1, 0, "Intel 82559ER Embedded 10/100 Ethernet" },
{ 0x1229, 0x01, 0, "Intel 82557 Pro/100 Ethernet" },
{ 0x1229, 0x02, 0, "Intel 82557 Pro/100 Ethernet" },
{ 0x1229, 0x03, 0, "Intel 82557 Pro/100 Ethernet" },
{ 0x1229, 0x04, 0, "Intel 82558 Pro/100 Ethernet" },
{ 0x1229, 0x05, 0, "Intel 82558 Pro/100 Ethernet" },
{ 0x1229, 0x06, 0, "Intel 82559 Pro/100 Ethernet" },
{ 0x1229, 0x07, 0, "Intel 82559 Pro/100 Ethernet" },
{ 0x1229, 0x08, 0, "Intel 82559 Pro/100 Ethernet" },
{ 0x1229, 0x09, 0, "Intel 82559ER Pro/100 Ethernet" },
{ 0x1229, 0x0c, 0, "Intel 82550 Pro/100 Ethernet" },
{ 0x1229, 0x0d, 0, "Intel 82550 Pro/100 Ethernet" },
{ 0x1229, 0x0e, 0, "Intel 82550 Pro/100 Ethernet" },
{ 0x1229, 0x0f, 0, "Intel 82551 Pro/100 Ethernet" },
{ 0x1229, 0x10, 0, "Intel 82551 Pro/100 Ethernet" },
{ 0x1229, -1, 0, "Intel 82557/8/9 Pro/100 Ethernet" },
{ 0x2449, -1, 2, "Intel 82801BA/CAM (ICH2/3) Pro/100 Ethernet" },
{ 0x27dc, -1, 7, "Intel 82801GB (ICH7) 10/100 Ethernet" },
{ 0, -1, 0, NULL },
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
};
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
#ifdef FXP_IP_CSUM_WAR
#define FXP_CSUM_FEATURES (CSUM_IP | CSUM_TCP | CSUM_UDP)
#else
#define FXP_CSUM_FEATURES (CSUM_TCP | CSUM_UDP)
#endif
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static int fxp_probe(device_t dev);
static int fxp_attach(device_t dev);
static int fxp_detach(device_t dev);
static int fxp_shutdown(device_t dev);
static int fxp_suspend(device_t dev);
static int fxp_resume(device_t dev);
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
static struct fxp_ident *fxp_find_ident(device_t dev);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static void fxp_intr(void *xsc);
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
static void fxp_rxcsum(struct fxp_softc *sc, struct ifnet *ifp,
struct mbuf *m, uint16_t status, int pos);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int fxp_intr_body(struct fxp_softc *sc, struct ifnet *ifp,
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint8_t statack, int count);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static void fxp_init(void *xsc);
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
static void fxp_init_body(struct fxp_softc *sc);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static void fxp_tick(void *xsc);
static void fxp_start(struct ifnet *ifp);
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
static void fxp_start_body(struct ifnet *ifp);
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
static int fxp_encap(struct fxp_softc *sc, struct mbuf **m_head);
static void fxp_txeof(struct fxp_softc *sc);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static void fxp_stop(struct fxp_softc *sc);
static void fxp_release(struct fxp_softc *sc);
static int fxp_ioctl(struct ifnet *ifp, u_long command,
caddr_t data);
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on fxp_tick() callout.
2006-11-30 14:58:01 +00:00
static void fxp_watchdog(struct fxp_softc *sc);
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
static void fxp_add_rfabuf(struct fxp_softc *sc,
struct fxp_rx *rxp);
static void fxp_discard_rfabuf(struct fxp_softc *sc,
struct fxp_rx *rxp);
static int fxp_new_rfabuf(struct fxp_softc *sc,
struct fxp_rx *rxp);
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
static int fxp_mc_addrs(struct fxp_softc *sc);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static void fxp_mc_setup(struct fxp_softc *sc);
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
static uint16_t fxp_eeprom_getword(struct fxp_softc *sc, int offset,
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
int autosize);
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
static void fxp_eeprom_putword(struct fxp_softc *sc, int offset,
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint16_t data);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static void fxp_autosize_eeprom(struct fxp_softc *sc);
static void fxp_read_eeprom(struct fxp_softc *sc, u_short *data,
int offset, int words);
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
static void fxp_write_eeprom(struct fxp_softc *sc, u_short *data,
int offset, int words);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static int fxp_ifmedia_upd(struct ifnet *ifp);
static void fxp_ifmedia_sts(struct ifnet *ifp,
struct ifmediareq *ifmr);
static int fxp_serial_ifmedia_upd(struct ifnet *ifp);
static void fxp_serial_ifmedia_sts(struct ifnet *ifp,
struct ifmediareq *ifmr);
Remove the volatile qualifier to apply to fxp_miibus_readreg().
2007-05-30 03:46:04 +00:00
static int fxp_miibus_readreg(device_t dev, int phy, int reg);
writereg returns an int.
2009-02-05 19:33:35 +00:00
static int fxp_miibus_writereg(device_t dev, int phy, int reg,
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
int value);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
static void fxp_load_ucode(struct fxp_softc *sc);
Export hardware MAC statistics through sysctl node. Previously fxp(4) already used to extract most hardware MAC statistics but it didn't show them. With this change, all MAC statistics counters are exported. Because there are a couple of new counters for 82558 and 82559, enable extended MAC statistics functionality to get these counters. Accoring to public data sheet, 82559 MAC statistics return 24 DWORD counters(3 counters are unknown at this moment) so increase MAC counter structure to meet the MAC statistics block size. The completion of MAC counter dump is now checked against FXP_STATS_DR_COMPLETE status code which is appended at the end of status block. Previously fxp(4) ignored the status of the FXP_SCB_COMMAND_CU_DUMPRESET command. fxp(4) does not wait for the completion of pending command before issuing FXP_SCB_COMMAND_CU_DUMPRESET. Instead it skips the command and try it next time. This scheme may show better performance but there is chance to loose updated counters after stopping controller. So make sure to update MAC statistics in fxp_stop(). While I'm here move sysctl node creation to fxp_sysctl_node(). Tested by: Larry Baird < lab <> gta dot com >
2010-05-09 22:16:15 +00:00
static void fxp_update_stats(struct fxp_softc *sc);
static void fxp_sysctl_node(struct fxp_softc *sc);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
static int sysctl_int_range(SYSCTL_HANDLER_ARGS,
int low, int high);
static int sysctl_hw_fxp_bundle_max(SYSCTL_HANDLER_ARGS);
static int sysctl_hw_fxp_int_delay(SYSCTL_HANDLER_ARGS);
De-inline some functions. It doesn't gain us anything and bloats code size by 3616 bytes. Furthemore, it was previously ignored by GCC. While I'm at it, fix some bogus comments.
2003-07-22 15:50:54 +00:00
static void fxp_scb_wait(struct fxp_softc *sc);
static void fxp_scb_cmd(struct fxp_softc *sc, int cmd);
static void fxp_dma_wait(struct fxp_softc *sc,
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
volatile uint16_t *status, bus_dma_tag_t dmat,
- Pass a busdma tag and a busdma map to fxp_dma_wait() so that it can do the necessary bus_dmamap_sync() calls. - While I'm here, reorder arguments to match other functions.
2003-04-08 18:56:45 +00:00
bus_dmamap_t map);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static device_method_t fxp_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, fxp_probe),
DEVMETHOD(device_attach, fxp_attach),
DEVMETHOD(device_detach, fxp_detach),
DEVMETHOD(device_shutdown, fxp_shutdown),
DEVMETHOD(device_suspend, fxp_suspend),
DEVMETHOD(device_resume, fxp_resume),
Implemented a better, dynamic, mechanism for adjusting the transmit threshold.
1996-09-19 09:15:20 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/* MII interface */
DEVMETHOD(miibus_readreg, fxp_miibus_readreg),
DEVMETHOD(miibus_writereg, fxp_miibus_writereg),
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
{ 0, 0 }
};
Implemented a more sophisticated mechanism for handling transmitter interrupts which now defers them until the transmit queue if filled up with completed buffers. This has two advantages: first, it reduces the number of transmitter interrupts to just 1/120th of the rate that they occured previously, and two, running down many buffers at once has much improved cache effects.
1998-08-04 08:53:12 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static driver_t fxp_driver = {
"fxp",
fxp_methods,
sizeof(struct fxp_softc),
};
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static devclass_t fxp_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(fxp, pci, fxp_driver, fxp_devclass, 0, 0);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
DRIVER_MODULE(miibus, fxp, miibus_driver, miibus_devclass, 0, 0);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies the resource allocation code significantly.
2005-09-27 09:01:11 +00:00
static struct resource_spec fxp_res_spec_mem[] = {
{ SYS_RES_MEMORY, FXP_PCI_MMBA, RF_ACTIVE },
{ SYS_RES_IRQ, 0, RF_ACTIVE | RF_SHAREABLE },
{ -1, 0 }
};
static struct resource_spec fxp_res_spec_io[] = {
{ SYS_RES_IOPORT, FXP_PCI_IOBA, RF_ACTIVE },
{ SYS_RES_IRQ, 0, RF_ACTIVE | RF_SHAREABLE },
{ -1, 0 }
};
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
/*
* Wait for the previous command to be accepted (but not necessarily
* completed).
*/
De-inline some functions. It doesn't gain us anything and bloats code size by 3616 bytes. Furthemore, it was previously ignored by GCC. While I'm at it, fix some bogus comments.
2003-07-22 15:50:54 +00:00
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_scb_wait(struct fxp_softc *sc)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Fix an unaligned I/O memory access in the event that a SCB times out. The FXP_SCR_FLOWCONTROL registers is at offset 0x19, but 2 bytes wide. It cannot be read as a word without causing a panic on architectures that enforce strict alignment. MFC after: 3 days
2005-09-21 04:36:40 +00:00
union {
uint16_t w;
uint8_t b[2];
} flowctl;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
int i = 10000;
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
while (CSR_READ_1(sc, FXP_CSR_SCB_COMMAND) && --i)
DELAY(2);
Fix an unaligned I/O memory access in the event that a SCB times out. The FXP_SCR_FLOWCONTROL registers is at offset 0x19, but 2 bytes wide. It cannot be read as a word without causing a panic on architectures that enforce strict alignment. MFC after: 3 days
2005-09-21 04:36:40 +00:00
if (i == 0) {
flowctl.b[0] = CSR_READ_1(sc, FXP_CSR_FLOWCONTROL);
flowctl.b[1] = CSR_READ_1(sc, FXP_CSR_FLOWCONTROL + 1);
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
device_printf(sc->dev, "SCB timeout: 0x%x 0x%x 0x%x 0x%x\n",
Add few cosmetic style fixes, and some debug information for SCB timeouts. Add VLAN support, obtained from Pedro J. Lobo (through Mike Tancsa).
2001-05-13 00:03:39 +00:00
CSR_READ_1(sc, FXP_CSR_SCB_COMMAND),
CSR_READ_1(sc, FXP_CSR_SCB_STATACK),
Fix an unaligned I/O memory access in the event that a SCB times out. The FXP_SCR_FLOWCONTROL registers is at offset 0x19, but 2 bytes wide. It cannot be read as a word without causing a panic on architectures that enforce strict alignment. MFC after: 3 days
2005-09-21 04:36:40 +00:00
CSR_READ_1(sc, FXP_CSR_SCB_RUSCUS), flowctl.w);
}
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
}
De-inline some functions. It doesn't gain us anything and bloats code size by 3616 bytes. Furthemore, it was previously ignored by GCC. While I'm at it, fix some bogus comments.
2003-07-22 15:50:54 +00:00
static void
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(struct fxp_softc *sc, int cmd)
{
if (cmd == FXP_SCB_COMMAND_CU_RESUME && sc->cu_resume_bug) {
CSR_WRITE_1(sc, FXP_CSR_SCB_COMMAND, FXP_CB_COMMAND_NOP);
fxp_scb_wait(sc);
}
CSR_WRITE_1(sc, FXP_CSR_SCB_COMMAND, cmd);
}
De-inline some functions. It doesn't gain us anything and bloats code size by 3616 bytes. Furthemore, it was previously ignored by GCC. While I'm at it, fix some bogus comments.
2003-07-22 15:50:54 +00:00
static void
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
fxp_dma_wait(struct fxp_softc *sc, volatile uint16_t *status,
- Pass a busdma tag and a busdma map to fxp_dma_wait() so that it can do the necessary bus_dmamap_sync() calls. - While I'm here, reorder arguments to match other functions.
2003-04-08 18:56:45 +00:00
bus_dma_tag_t dmat, bus_dmamap_t map)
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
{
Controller will dma SCB command status for a given command and driver should read updated status back after issuing a SCB command. To send a command to controller and read updated status back, driver should synchronize both memory read and write operations with device. Fix bus_dmamap_sync operation specifier used in fxp_dma_wait() by adding both memory read and memory write operations.
2009-06-10 01:15:30 +00:00
int i;
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
Controller will dma SCB command status for a given command and driver should read updated status back after issuing a SCB command. To send a command to controller and read updated status back, driver should synchronize both memory read and write operations with device. Fix bus_dmamap_sync operation specifier used in fxp_dma_wait() by adding both memory read and memory write operations.
2009-06-10 01:15:30 +00:00
for (i = 10000; i > 0; i--) {
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
DELAY(2);
Controller will dma SCB command status for a given command and driver should read updated status back after issuing a SCB command. To send a command to controller and read updated status back, driver should synchronize both memory read and write operations with device. Fix bus_dmamap_sync operation specifier used in fxp_dma_wait() by adding both memory read and memory write operations.
2009-06-10 01:15:30 +00:00
bus_dmamap_sync(dmat, map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
if ((le16toh(*status) & FXP_CB_STATUS_C) != 0)
break;
- Pass a busdma tag and a busdma map to fxp_dma_wait() so that it can do the necessary bus_dmamap_sync() calls. - While I'm here, reorder arguments to match other functions.
2003-04-08 18:56:45 +00:00
}
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
if (i == 0)
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
device_printf(sc->dev, "DMA timeout\n");
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
}
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
static struct fxp_ident *
fxp_find_ident(device_t dev)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint16_t devid;
uint8_t revid;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
struct fxp_ident *ident;
Added support for the i82559ER (10/100Mbps NIC for embedded applications). Product device ID provided by: Les Biffle <les@ns3.safety.net>
2000-06-18 10:26:09 +00:00
if (pci_get_vendor(dev) == FXP_VENDORID_INTEL) {
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
devid = pci_get_device(dev);
Use the revision ID from PCI configuration space to identify Intel 8255x chips more precisely. The information was obtained from Intel's Open Source Software Developer Manual for the 8255x. MFC after: 1 day
2003-08-04 00:17:16 +00:00
revid = pci_get_revid(dev);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
for (ident = fxp_ident_table; ident->name != NULL; ident++) {
Use the revision ID from PCI configuration space to identify Intel 8255x chips more precisely. The information was obtained from Intel's Open Source Software Developer Manual for the 8255x. MFC after: 1 day
2003-08-04 00:17:16 +00:00
if (ident->devid == devid &&
(ident->revid == revid || ident->revid == -1)) {
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
return (ident);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
}
Added support for the i82559ER (10/100Mbps NIC for embedded applications). Product device ID provided by: Les Biffle <les@ns3.safety.net>
2000-06-18 10:26:09 +00:00
}
Recognise the new 82559 chip id as used on the InBusiness 10/100 adapter. I have an 82559 card with the same id as the other 8255[78] chips, but that was made with a date code of 0699 (June 99). The submitter shows this working with the probe etc, but doesn't actually say it works as on the ethernet. :-) Assuming it does, this is a RELENG_3 merge candidate. Submitted by: Steven E Lumos <slumos@sam.ISRI.UNLV.EDU>
1999-09-06 06:15:18 +00:00
}
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
return (NULL);
}
/*
* Return identification string if this device is ours.
*/
static int
fxp_probe(device_t dev)
{
struct fxp_ident *ident;
ident = fxp_find_ident(dev);
if (ident != NULL) {
device_set_desc(dev, ident->name);
return (BUS_PROBE_DEFAULT);
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
return (ENXIO);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
static void
fxp_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint32_t *addr;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (error)
return;
KASSERT(nseg == 1, ("too many DMA segments, %d should be 1", nseg));
addr = arg;
*addr = segs->ds_addr;
}
Bring the 'new-bus' to the i386. This extensively changes the way the i386 platform boots, it is no longer ISA-centric, and is fully dynamic. Most old drivers compile and run without modification via 'compatability shims' to enable a smoother transition. eisa, isapnp and pccard* are not yet using the new resource manager. Once fully converted, all drivers will be loadable, including PCI and ISA. (Some other changes appear to have snuck in, including a port of Soren's ATA driver to the Alpha. Soren, back this out if you need to.) This is a checkpoint of work-in-progress, but is quite functional. The bulk of the work was done over the last few years by Doug Rabson and Garrett Wollman. Approved by: core
1999-04-16 21:22:55 +00:00
static int
fxp_attach(device_t dev)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Fix the panic recently reported on -current@ occuring when configuring a vlan interface attached to a fxp(4) card when it has not been initialized yet. We now set the links from our internel TX descriptor structure to the TX command blocks at attach time rather than at init time. While I'm here, slightly improve the style in fxp_attach(). PR: kern/78112 Reported by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> and others Tested by: flz, Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> MFC after: 1 week
2005-03-03 00:22:59 +00:00
struct fxp_softc *sc;
struct fxp_cb_tx *tcbp;
struct fxp_tx *txp;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
struct fxp_rx *rxp;
Fix the panic recently reported on -current@ occuring when configuring a vlan interface attached to a fxp(4) card when it has not been initialized yet. We now set the links from our internel TX descriptor structure to the TX command blocks at attach time rather than at init time. While I'm here, slightly improve the style in fxp_attach(). PR: kern/78112 Reported by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> and others Tested by: flz, Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> MFC after: 1 week
2005-03-03 00:22:59 +00:00
struct ifnet *ifp;
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint32_t val;
uint16_t data, myea[ETHER_ADDR_LEN / 2];
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
u_char eaddr[ETHER_ADDR_LEN];
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
int i, pmc, prefer_iomap;
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
int error;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Fix the panic recently reported on -current@ occuring when configuring a vlan interface attached to a fxp(4) card when it has not been initialized yet. We now set the links from our internel TX descriptor structure to the TX command blocks at attach time rather than at init time. While I'm here, slightly improve the style in fxp_attach(). PR: kern/78112 Reported by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> and others Tested by: flz, Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> MFC after: 1 week
2005-03-03 00:22:59 +00:00
error = 0;
sc = device_get_softc(dev);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
sc->dev = dev;
Change callers of mtx_init() to pass in an appropriate lock type name. In most cases NULL is passed, but in some cases such as network driver locks (which use the MTX_NETWORK_LOCK macro) and UMA zone locks, a name is used. Tested on: i386, alpha, sparc64
2002-04-04 21:03:38 +00:00
mtx_init(&sc->sc_mtx, device_get_nameunit(dev), MTX_NETWORK_LOCK,
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
MTX_DEF);
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
callout_init_mtx(&sc->stat_ch, &sc->sc_mtx, 0);
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
ifmedia_init(&sc->sc_media, 0, fxp_serial_ifmedia_upd,
fxp_serial_ifmedia_sts);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Move if_alloc() up in fxp_attach() so there's an ifp before mii_phy_probe() is called. Committed via: fxp0
2005-06-11 00:47:34 +00:00
ifp = sc->ifp = if_alloc(IFT_ETHER);
if (ifp == NULL) {
device_printf(dev, "can not if_alloc()\n");
error = ENOSPC;
goto fail;
}
Turn on PCI bus mastering in driver attach routine to avoid hanging the system if the BIOS has not initialized the interface. PR: 10150 Reviewed by: dg
1999-03-20 04:51:25 +00:00
/*
Update comment to match reality.
2003-04-16 09:16:55 +00:00
* Enable bus mastering.
Turn on PCI bus mastering in driver attach routine to avoid hanging the system if the BIOS has not initialized the interface. PR: 10150 Reviewed by: dg
1999-03-20 04:51:25 +00:00
*/
Use pci_enable_{busmaster|io}() instead of directly writing into PCI configuration register. Submitted by: mdodd
2003-04-14 15:50:04 +00:00
pci_enable_busmaster(dev);
Allow fxp to configure in I/O space if the user wants it and specifies an override as a loader settable variable (fxp_iomap). fxp_iomap is a bitmap of fxp units that should be configured to use PCI I/O space in stead of PCI Memory space. Reviewed by: Kees Jan Koster <dutchman@tccn.cs.kun.nl>, dg@freebsd.org
2001-01-23 23:22:17 +00:00
val = pci_read_config(dev, PCIR_COMMAND, 2);
Remove the setting of the pci config variables on power state changes. The bus does this now.
2004-06-28 20:26:21 +00:00
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
/*
Allow fxp to configure in I/O space if the user wants it and specifies an override as a loader settable variable (fxp_iomap). fxp_iomap is a bitmap of fxp units that should be configured to use PCI I/O space in stead of PCI Memory space. Reviewed by: Kees Jan Koster <dutchman@tccn.cs.kun.nl>, dg@freebsd.org
2001-01-23 23:22:17 +00:00
* Figure out which we should try first - memory mapping or i/o mapping?
* We default to memory mapping. Then we accept an override from the
* command line. Then we check to see which one is enabled.
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
*/
A better mousetrap: use device hints, as in: hint.fxp.0.prefer_iomap="1" to set IO vs. Memory space mapping.
2001-02-27 22:57:32 +00:00
prefer_iomap = 0;
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies the resource allocation code significantly.
2005-09-27 09:01:11 +00:00
resource_int_value(device_get_name(dev), device_get_unit(dev),
"prefer_iomap", &prefer_iomap);
if (prefer_iomap)
sc->fxp_spec = fxp_res_spec_io;
else
sc->fxp_spec = fxp_res_spec_mem;
Allow fxp to configure in I/O space if the user wants it and specifies an override as a loader settable variable (fxp_iomap). fxp_iomap is a bitmap of fxp units that should be configured to use PCI I/O space in stead of PCI Memory space. Reviewed by: Kees Jan Koster <dutchman@tccn.cs.kun.nl>, dg@freebsd.org
2001-01-23 23:22:17 +00:00
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies the resource allocation code significantly.
2005-09-27 09:01:11 +00:00
error = bus_alloc_resources(dev, sc->fxp_spec, sc->fxp_res);
if (error) {
if (sc->fxp_spec == fxp_res_spec_mem)
sc->fxp_spec = fxp_res_spec_io;
else
sc->fxp_spec = fxp_res_spec_mem;
error = bus_alloc_resources(dev, sc->fxp_spec, sc->fxp_res);
Allow fxp to configure in I/O space if the user wants it and specifies an override as a loader settable variable (fxp_iomap). fxp_iomap is a bitmap of fxp units that should be configured to use PCI I/O space in stead of PCI Memory space. Reviewed by: Kees Jan Koster <dutchman@tccn.cs.kun.nl>, dg@freebsd.org
2001-01-23 23:22:17 +00:00
}
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies the resource allocation code significantly.
2005-09-27 09:01:11 +00:00
if (error) {
device_printf(dev, "could not allocate resources\n");
Bring the 'new-bus' to the i386. This extensively changes the way the i386 platform boots, it is no longer ISA-centric, and is fully dynamic. Most old drivers compile and run without modification via 'compatability shims' to enable a smoother transition. eisa, isapnp and pccard* are not yet using the new resource manager. Once fully converted, all drivers will be loadable, including PCI and ISA. (Some other changes appear to have snuck in, including a port of Soren's ATA driver to the Alpha. Soren, back this out if you need to.) This is a checkpoint of work-in-progress, but is quite functional. The bulk of the work was done over the last few years by Doug Rabson and Garrett Wollman. Approved by: core
1999-04-16 21:22:55 +00:00
error = ENXIO;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
goto fail;
Allow fxp to configure in I/O space if the user wants it and specifies an override as a loader settable variable (fxp_iomap). fxp_iomap is a bitmap of fxp units that should be configured to use PCI I/O space in stead of PCI Memory space. Reviewed by: Kees Jan Koster <dutchman@tccn.cs.kun.nl>, dg@freebsd.org
2001-01-23 23:22:17 +00:00
}
Make the fxp driver work on alpha, rather than panic the machine on boot and/or when using the card. o Convert the driver to using bus_space. This allows alphas with fxp's to boot, rather than panic'ing because rman_get_virtual() doesn't really return a virtual address on alphas. o Fix an alpha unaligned access error caused by some misfeature of gcc/egcs: if link_addr & rbd_addr in the fxp_rfa struct are 32 bit quantities, egcs will assume they are naturally aligned. So it will do a ldl & some shifty/masky to twiddle 16 bit values in fxp_lwcopy(). However, if they are 16-bit aligned, the ldl will actually be done on a 16-bit aligned value & we will panic with an unaligned access error... Changing their definition to an array of chars seems to fix this. I obtained this from NetBSD. I've tested this on both i386 & alpha.
1999-09-30 19:03:12 +00:00
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies the resource allocation code significantly.
2005-09-27 09:01:11 +00:00
if (bootverbose) {
device_printf(dev, "using %s space register mapping\n",
sc->fxp_spec == fxp_res_spec_mem ? "memory" : "I/O");
Bring the 'new-bus' to the i386. This extensively changes the way the i386 platform boots, it is no longer ISA-centric, and is fully dynamic. Most old drivers compile and run without modification via 'compatability shims' to enable a smoother transition. eisa, isapnp and pccard* are not yet using the new resource manager. Once fully converted, all drivers will be loadable, including PCI and ISA. (Some other changes appear to have snuck in, including a port of Soren's ATA driver to the Alpha. Soren, back this out if you need to.) This is a checkpoint of work-in-progress, but is quite functional. The bulk of the work was done over the last few years by Doug Rabson and Garrett Wollman. Approved by: core
1999-04-16 21:22:55 +00:00
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/*
Due to possible PCI bus lock-up issues fxp(4) didn't perform full hardware reset in attach phase. Selective reset does not clear configured parameters so I think full hardware reset is required. To prevent PCI bus lock-up, do selective reset first which will get off the controller from PCI bus and request software reset after selective reset. Software reset will unmask interrupts so disable it after the reset.
2009-06-21 06:18:19 +00:00
* Put CU/RU idle state and prepare full reset.
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
*/
CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SELECTIVE_RESET);
DELAY(10);
Due to possible PCI bus lock-up issues fxp(4) didn't perform full hardware reset in attach phase. Selective reset does not clear configured parameters so I think full hardware reset is required. To prevent PCI bus lock-up, do selective reset first which will get off the controller from PCI bus and request software reset after selective reset. Software reset will unmask interrupts so disable it after the reset.
2009-06-21 06:18:19 +00:00
/* Full reset and disable interrupts. */
CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SOFTWARE_RESET);
DELAY(10);
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, FXP_SCB_INTR_DISABLE);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/*
* Find out how large of an SEEPROM we have.
*/
fxp_autosize_eeprom(sc);
Fix a stupid bogon from myself, sc->revision wasn't initialized when testing it to know whether we should enable the 82503 serial mode... Move code to the right location and disallow the use of the 82503 serial mode if the sc->revision field is 0 again. This makes fxp(4) work correctly with ATMEL 350 93C46 cards (3 port 82559 based with a 82555 PHY), as well as with the older ATMEL 220 93C46 (same flavour) and with the even older 10Mbps-only 82557 cards with the 82503 serial interface. Tested by: Andre Albsmeier <andrer@albsmeier.net>, krion MFC after: 2 weeks
2005-02-27 15:12:50 +00:00
/*
* Find out the chip revision; lump all 82557 revs together.
*/
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
sc->ident = fxp_find_ident(dev);
if (sc->ident->ich > 0) {
/* Assume ICH controllers are 82559. */
sc->revision = FXP_REV_82559_A0;
} else {
fxp_read_eeprom(sc, &data, 5, 1);
if ((data >> 8) == 1)
sc->revision = FXP_REV_82557;
else
sc->revision = pci_get_revid(dev);
}
Fix a stupid bogon from myself, sc->revision wasn't initialized when testing it to know whether we should enable the 82503 serial mode... Move code to the right location and disallow the use of the 82503 serial mode if the sc->revision field is 0 again. This makes fxp(4) work correctly with ATMEL 350 93C46 cards (3 port 82559 based with a 82555 PHY), as well as with the older ATMEL 220 93C46 (same flavour) and with the even older 10Mbps-only 82557 cards with the 82503 serial interface. Tested by: Andre Albsmeier <andrer@albsmeier.net>, krion MFC after: 2 weeks
2005-02-27 15:12:50 +00:00
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
/*
* Check availability of WOL. 82559ER does not support WOL.
*/
if (sc->revision >= FXP_REV_82558_A4 &&
sc->revision != FXP_REV_82559S_A) {
fxp_read_eeprom(sc, &data, 10, 1);
if ((data & 0x20) != 0 &&
pci_find_extcap(sc->dev, PCIY_PMG, &pmc) == 0)
sc->flags |= FXP_FLAG_WOLCAP;
}
Don't blindly enable Rx lock-up workaround. Newer chips do not need the Rx lock-up workaround. Obtained from: NetBSD
2009-06-21 06:27:35 +00:00
/* Receiver lock-up workaround detection. */
fxp_read_eeprom(sc, &data, 3, 1);
if ((data & 0x03) != 0x03) {
sc->flags |= FXP_FLAG_RXBUG;
device_printf(dev, "Enabling Rx lock-up workaround\n");
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/*
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
* Determine whether we must use the 503 serial interface.
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
*/
fxp_read_eeprom(sc, &data, 6, 1);
Fix a stupid bogon from myself, sc->revision wasn't initialized when testing it to know whether we should enable the 82503 serial mode... Move code to the right location and disallow the use of the 82503 serial mode if the sc->revision field is 0 again. This makes fxp(4) work correctly with ATMEL 350 93C46 cards (3 port 82559 based with a 82555 PHY), as well as with the older ATMEL 220 93C46 (same flavour) and with the even older 10Mbps-only 82557 cards with the 82503 serial interface. Tested by: Andre Albsmeier <andrer@albsmeier.net>, krion MFC after: 2 weeks
2005-02-27 15:12:50 +00:00
if (sc->revision == FXP_REV_82557 && (data & FXP_PHY_DEVICE_MASK) != 0
Only try to use the 82503 serial interface for the 82557 chipsets. The datasheet says it is only valid for such chipsets and shouldn't be used with others. This fixes some 82559 based cards which otherwise only work at 10Mbit. MFC after: 5 days Tested by: krion
2004-12-20 10:18:21 +00:00
&& (data & FXP_PHY_SERIAL_ONLY))
Use " |= " to enable special media handling for fxp with no MII, instead of " &= ". Also change the MII PHY device mask to check the correct bits. Cookie to: Andre Albsmeier <andre.albsmeier@mchp.siemens.de> Pointy hat to: me
2001-05-15 18:52:40 +00:00
sc->flags |= FXP_FLAG_SERIAL_MEDIA;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
Export hardware MAC statistics through sysctl node. Previously fxp(4) already used to extract most hardware MAC statistics but it didn't show them. With this change, all MAC statistics counters are exported. Because there are a couple of new counters for 82558 and 82559, enable extended MAC statistics functionality to get these counters. Accoring to public data sheet, 82559 MAC statistics return 24 DWORD counters(3 counters are unknown at this moment) so increase MAC counter structure to meet the MAC statistics block size. The completion of MAC counter dump is now checked against FXP_STATS_DR_COMPLETE status code which is appended at the end of status block. Previously fxp(4) ignored the status of the FXP_SCB_COMMAND_CU_DUMPRESET command. fxp(4) does not wait for the completion of pending command before issuing FXP_SCB_COMMAND_CU_DUMPRESET. Instead it skips the command and try it next time. This scheme may show better performance but there is chance to loose updated counters after stopping controller. So make sure to update MAC statistics in fxp_stop(). While I'm here move sysctl node creation to fxp_sysctl_node(). Tested by: Larry Baird < lab <> gta dot com >
2010-05-09 22:16:15 +00:00
fxp_sysctl_node(sc);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
/*
* Enable workarounds for certain chip revision deficiencies.
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
*
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
* Systems based on the ICH2/ICH2-M chip from Intel, and possibly
* some systems based a normal 82559 design, have a defect where
* the chip can cause a PCI protocol violation if it receives
Whitespace fix.
2008-11-25 00:48:05 +00:00
* a CU_RESUME command when it is entering the IDLE state. The
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
* workaround is to disable Dynamic Standby Mode, so the chip never
* deasserts CLKRUN#, and always remains in an active state.
*
* See Intel 82801BA/82801BAM Specification Update, Errata #30.
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
*/
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
if ((sc->ident->ich >= 2 && sc->ident->ich <= 3) ||
(sc->ident->ich == 0 && sc->revision >= FXP_REV_82559_A0)) {
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
fxp_read_eeprom(sc, &data, 10, 1);
if (data & 0x02) { /* STB enable */
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint16_t cksum;
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
int i;
device_printf(dev,
Tone down and remove some obnoxious warnings that are slightly overkill.
2001-12-13 16:13:31 +00:00
"Disabling dynamic standby mode in EEPROM\n");
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
data &= ~0x02;
fxp_write_eeprom(sc, &data, 10, 1);
device_printf(dev, "New EEPROM ID: 0x%x\n", data);
cksum = 0;
for (i = 0; i < (1 << sc->eeprom_size) - 1; i++) {
fxp_read_eeprom(sc, &data, i, 1);
cksum += data;
}
i = (1 << sc->eeprom_size) - 1;
cksum = 0xBABA - cksum;
fxp_read_eeprom(sc, &data, i, 1);
fxp_write_eeprom(sc, &cksum, i, 1);
device_printf(dev,
"EEPROM checksum @ 0x%x: 0x%x -> 0x%x\n",
i, data, cksum);
#if 1
/*
* If the user elects to continue, try the software
* workaround, as it is better than nothing.
*/
sc->flags |= FXP_FLAG_CU_RESUME_BUG;
#endif
}
}
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
/*
* If we are not a 82557 chip, we can enable extended features.
*/
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
if (sc->revision != FXP_REV_82557) {
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
/*
Only turn on MWI if the PCI configuration word indicates that it is supported, in addition to checking for a valid cacheline size. Add a missing splx() in fxp_tick that got dropped. Found by: peter MFC in: 3 days
2001-07-19 15:48:00 +00:00
* If MWI is enabled in the PCI configuration, and there
* is a valid cacheline size (8 or 16 dwords), then tell
* the board to turn on MWI.
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
*/
Only turn on MWI if the PCI configuration word indicates that it is supported, in addition to checking for a valid cacheline size. Add a missing splx() in fxp_tick that got dropped. Found by: peter MFC in: 3 days
2001-07-19 15:48:00 +00:00
if (val & PCIM_CMD_MWRICEN &&
pci_read_config(dev, PCIR_CACHELNSZ, 1) != 0)
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
sc->flags |= FXP_FLAG_MWI_ENABLE;
/* turn on the extended TxCB feature */
sc->flags |= FXP_FLAG_EXT_TXCB;
POLA dictates that VLAN_MTU be enabled by default. In particular, disabling it was likely to break configurations involving ng_vlan(4) since the latter couldn't control the parent's VLAN_MTU in the way vlan(4) did. Pointed out by: ru
2004-05-27 14:36:58 +00:00
/* enable reception of long frames for VLAN */
sc->flags |= FXP_FLAG_LONG_PKT_EN;
} else {
/* a hack to get long VLAN frames on a 82557 */
sc->flags |= FXP_FLAG_SAVE_BAD;
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
}
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
/* For 82559 or later chips, Rx checksum offload is supported. */
It seems some 82559ER controllers do not support Rx checksum offloading. Datasheet said nothing about the limitation of 82559ER except WOL. Explicitly disable Rx checksum offloading for controllers that is known to lack the capability. PR: kern/138135 Tested by: Gooderum, Mark < mgooderum <> websense dot com >
2009-09-28 19:40:16 +00:00
if (sc->revision >= FXP_REV_82559_A0) {
/* 82559ER does not support Rx checksum offloading. */
if (sc->ident->devid != 0x1209)
sc->flags |= FXP_FLAG_82559_RXCSUM;
}
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
/*
* Enable use of extended RFDs and TCBs for 82550
* and later chips. Note: we need extended TXCB support
* too, but that's already enabled by the code above.
* Be careful to do this only on the right devices.
*/
Be more conservative when enabling extended features. There are fxp(4) NICs out there that have an utterly bogus revision ID. Reported by: Denis Shaposhnikov <dsh@vlink.ru>
2005-04-22 13:05:53 +00:00
if (sc->revision == FXP_REV_82550 || sc->revision == FXP_REV_82550_C ||
sc->revision == FXP_REV_82551_E || sc->revision == FXP_REV_82551_F
|| sc->revision == FXP_REV_82551_10) {
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
sc->rfa_size = sizeof (struct fxp_rfa);
sc->tx_cmd = FXP_CB_COMMAND_IPCBXMIT;
sc->flags |= FXP_FLAG_EXT_RFA;
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
/* Use extended RFA instead of 82559 checksum mode. */
sc->flags &= ~FXP_FLAG_82559_RXCSUM;
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
} else {
sc->rfa_size = sizeof (struct fxp_rfa) - FXP_RFAX_LEN;
sc->tx_cmd = FXP_CB_COMMAND_XMIT;
}
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
/*
* Allocate DMA tags and DMA safe memory.
*/
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
sc->maxtxseg = FXP_NTXSEG;
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
sc->maxsegsize = MCLBYTES;
if (sc->flags & FXP_FLAG_EXT_RFA) {
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
sc->maxtxseg--;
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
sc->maxsegsize = FXP_TSO_SEGSIZE;
}
Change the remainder of the drivers for DMA'ing devices enabled in the sparc64 GENERIC and the sound device drivers known working on sparc64 to use bus_get_dma_tag() to obtain the parent DMA tag so we can get rid of the sparc64_root_dma_tag kludge eventually. Except for ath(4), sk(4), stge(4) and ti(4) these changes are runtime tested (unless I booted up the wrong kernels again...).
2007-01-21 19:32:51 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(dev), 2, 0,
BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
sc->maxsegsize * sc->maxtxseg + sizeof(struct ether_vlan_header),
sc->maxtxseg, sc->maxsegsize, 0,
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
busdma_lock_mutex, &Giant, &sc->fxp_txmtag);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (error) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
device_printf(dev, "could not create TX DMA tag\n");
goto fail;
}
error = bus_dma_tag_create(bus_get_dma_tag(dev), 2, 0,
BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
MCLBYTES, 1, MCLBYTES, 0,
busdma_lock_mutex, &Giant, &sc->fxp_rxmtag);
if (error) {
device_printf(dev, "could not create RX DMA tag\n");
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
goto fail;
}
Change the remainder of the drivers for DMA'ing devices enabled in the sparc64 GENERIC and the sound device drivers known working on sparc64 to use bus_get_dma_tag() to obtain the parent DMA tag so we can get rid of the sparc64_root_dma_tag kludge eventually. Except for ath(4), sk(4), stge(4) and ti(4) these changes are runtime tested (unless I booted up the wrong kernels again...).
2007-01-21 19:32:51 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(dev), 4, 0,
BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
sizeof(struct fxp_stats), 1, sizeof(struct fxp_stats), 0,
busdma_lock_mutex, &Giant, &sc->fxp_stag);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (error) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
device_printf(dev, "could not create stats DMA tag\n");
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
goto fail;
}
error = bus_dmamem_alloc(sc->fxp_stag, (void **)&sc->fxp_stats,
Use the BUS_DMA_ZERO flag.
2003-07-27 14:00:02 +00:00
BUS_DMA_NOWAIT | BUS_DMA_ZERO, &sc->fxp_smap);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
if (error) {
device_printf(dev, "could not allocate stats DMA memory\n");
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
goto fail;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
}
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
error = bus_dmamap_load(sc->fxp_stag, sc->fxp_smap, sc->fxp_stats,
sizeof(struct fxp_stats), fxp_dma_map_addr, &sc->stats_addr, 0);
if (error) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
device_printf(dev, "could not load the stats DMA buffer\n");
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
goto fail;
}
Change the remainder of the drivers for DMA'ing devices enabled in the sparc64 GENERIC and the sound device drivers known working on sparc64 to use bus_get_dma_tag() to obtain the parent DMA tag so we can get rid of the sparc64_root_dma_tag kludge eventually. Except for ath(4), sk(4), stge(4) and ti(4) these changes are runtime tested (unless I booted up the wrong kernels again...).
2007-01-21 19:32:51 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(dev), 4, 0,
BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
FXP_TXCB_SZ, 1, FXP_TXCB_SZ, 0,
busdma_lock_mutex, &Giant, &sc->cbl_tag);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (error) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
device_printf(dev, "could not create TxCB DMA tag\n");
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
goto fail;
}
error = bus_dmamem_alloc(sc->cbl_tag, (void **)&sc->fxp_desc.cbl_list,
Use the BUS_DMA_ZERO flag.
2003-07-27 14:00:02 +00:00
BUS_DMA_NOWAIT | BUS_DMA_ZERO, &sc->cbl_map);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
if (error) {
device_printf(dev, "could not allocate TxCB DMA memory\n");
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
goto fail;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
}
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
error = bus_dmamap_load(sc->cbl_tag, sc->cbl_map,
sc->fxp_desc.cbl_list, FXP_TXCB_SZ, fxp_dma_map_addr,
&sc->fxp_desc.cbl_addr, 0);
if (error) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
device_printf(dev, "could not load TxCB DMA buffer\n");
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
goto fail;
}
Change the remainder of the drivers for DMA'ing devices enabled in the sparc64 GENERIC and the sound device drivers known working on sparc64 to use bus_get_dma_tag() to obtain the parent DMA tag so we can get rid of the sparc64_root_dma_tag kludge eventually. Except for ath(4), sk(4), stge(4) and ti(4) these changes are runtime tested (unless I booted up the wrong kernels again...).
2007-01-21 19:32:51 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(dev), 4, 0,
BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL,
sizeof(struct fxp_cb_mcs), 1, sizeof(struct fxp_cb_mcs), 0,
busdma_lock_mutex, &Giant, &sc->mcs_tag);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (error) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
device_printf(dev,
"could not create multicast setup DMA tag\n");
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
goto fail;
}
error = bus_dmamem_alloc(sc->mcs_tag, (void **)&sc->mcsp,
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
BUS_DMA_NOWAIT | BUS_DMA_ZERO, &sc->mcs_map);
if (error) {
device_printf(dev,
"could not allocate multicast setup DMA memory\n");
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
goto fail;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
}
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
error = bus_dmamap_load(sc->mcs_tag, sc->mcs_map, sc->mcsp,
sizeof(struct fxp_cb_mcs), fxp_dma_map_addr, &sc->mcs_addr, 0);
if (error) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
device_printf(dev,
"can't load the multicast setup DMA buffer\n");
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
goto fail;
}
/*
Fix the panic recently reported on -current@ occuring when configuring a vlan interface attached to a fxp(4) card when it has not been initialized yet. We now set the links from our internel TX descriptor structure to the TX command blocks at attach time rather than at init time. While I'm here, slightly improve the style in fxp_attach(). PR: kern/78112 Reported by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> and others Tested by: flz, Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> MFC after: 1 week
2005-03-03 00:22:59 +00:00
* Pre-allocate the TX DMA maps and setup the pointers to
* the TX command blocks.
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
*/
Fix the panic recently reported on -current@ occuring when configuring a vlan interface attached to a fxp(4) card when it has not been initialized yet. We now set the links from our internel TX descriptor structure to the TX command blocks at attach time rather than at init time. While I'm here, slightly improve the style in fxp_attach(). PR: kern/78112 Reported by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> and others Tested by: flz, Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> MFC after: 1 week
2005-03-03 00:22:59 +00:00
txp = sc->fxp_desc.tx_list;
tcbp = sc->fxp_desc.cbl_list;
- Change a FXP_NRFABUFS which would have been a FXP_NTXCB. - Correct some bus_dmamap_sync() calls. Submitted by: jake
2003-04-05 12:43:21 +00:00
for (i = 0; i < FXP_NTXCB; i++) {
Fix the panic recently reported on -current@ occuring when configuring a vlan interface attached to a fxp(4) card when it has not been initialized yet. We now set the links from our internel TX descriptor structure to the TX command blocks at attach time rather than at init time. While I'm here, slightly improve the style in fxp_attach(). PR: kern/78112 Reported by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> and others Tested by: flz, Gavin Atkinson <gavin.atkinson@ury.york.ac.uk> MFC after: 1 week
2005-03-03 00:22:59 +00:00
txp[i].tx_cb = tcbp + i;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
error = bus_dmamap_create(sc->fxp_txmtag, 0, &txp[i].tx_map);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (error) {
device_printf(dev, "can't create DMA map for TX\n");
goto fail;
}
}
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
error = bus_dmamap_create(sc->fxp_rxmtag, 0, &sc->spare_map);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (error) {
device_printf(dev, "can't create spare DMA map\n");
goto fail;
}
/*
* Pre-allocate our receive buffers.
*/
sc->fxp_desc.rx_head = sc->fxp_desc.rx_tail = NULL;
for (i = 0; i < FXP_NRFABUFS; i++) {
rxp = &sc->fxp_desc.rx_list[i];
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
error = bus_dmamap_create(sc->fxp_rxmtag, 0, &rxp->rx_map);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (error) {
device_printf(dev, "can't create DMA map for RX\n");
goto fail;
}
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
if (fxp_new_rfabuf(sc, rxp) != 0) {
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
error = ENOMEM;
goto fail;
}
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
fxp_add_rfabuf(sc, rxp);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/*
* Read MAC address.
*/
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
fxp_read_eeprom(sc, myea, 0, 3);
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
eaddr[0] = myea[0] & 0xff;
eaddr[1] = myea[0] >> 8;
eaddr[2] = myea[1] & 0xff;
eaddr[3] = myea[1] >> 8;
eaddr[4] = myea[2] & 0xff;
eaddr[5] = myea[2] >> 8;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
if (bootverbose) {
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
device_printf(dev, "PCI IDs: %04x %04x %04x %04x %04x\n",
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
pci_get_vendor(dev), pci_get_device(dev),
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
pci_get_subvendor(dev), pci_get_subdevice(dev),
pci_get_revid(dev));
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
fxp_read_eeprom(sc, &data, 10, 1);
device_printf(dev, "Dynamic Standby mode is %s\n",
data & 0x02 ? "enabled" : "disabled");
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
}
/*
* If this is only a 10Mbps device, then there is no MII, and
* the PHY will use a serial interface instead.
*
* The Seeq 80c24 AutoDUPLEX(tm) Ethernet Interface Adapter
* doesn't have a programming interface of any sort. The
* media is sensed automatically based on how the link partner
* is configured. This is, in essence, manual configuration.
*/
if (sc->flags & FXP_FLAG_SERIAL_MEDIA) {
ifmedia_add(&sc->sc_media, IFM_ETHER|IFM_MANUAL, 0, NULL);
ifmedia_set(&sc->sc_media, IFM_ETHER|IFM_MANUAL);
} else {
if (mii_phy_probe(dev, &sc->miibus, fxp_ifmedia_upd,
fxp_ifmedia_sts)) {
device_printf(dev, "MII without any PHY!\n");
error = ENXIO;
goto fail;
}
}
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
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));
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
ifp->if_init = fxp_init;
ifp->if_softc = sc;
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
ifp->if_ioctl = fxp_ioctl;
ifp->if_start = fxp_start;
Cosmetic: Set capability bits in a consistent way. Add a comment on why the VLAN_MTU stuff comes after ether_ifattach().
2004-05-21 20:34:04 +00:00
ifp->if_capabilities = ifp->if_capenable = 0;
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
/* Enable checksum offload/TSO for 82550 or better chips */
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
if (sc->flags & FXP_FLAG_EXT_RFA) {
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
ifp->if_hwassist = FXP_CSUM_FEATURES | CSUM_TSO;
ifp->if_capabilities |= IFCAP_HWCSUM | IFCAP_TSO4;
ifp->if_capenable |= IFCAP_HWCSUM | IFCAP_TSO4;
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
}
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
if (sc->flags & FXP_FLAG_82559_RXCSUM) {
ifp->if_capabilities |= IFCAP_RXCSUM;
ifp->if_capenable |= IFCAP_RXCSUM;
}
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
if (sc->flags & FXP_FLAG_WOLCAP) {
ifp->if_capabilities |= IFCAP_WOL_MAGIC;
ifp->if_capenable |= IFCAP_WOL_MAGIC;
}
Implemented per-interface polling(4) control.
2004-04-11 15:35:49 +00:00
#ifdef DEVICE_POLLING
/* Inform the world we support polling. */
ifp->if_capabilities |= IFCAP_POLLING;
#endif
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
/*
* Attach the interface.
*/
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ether_ifattach(ifp, eaddr);
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
Add few cosmetic style fixes, and some debug information for SCB timeouts. Add VLAN support, obtained from Pedro J. Lobo (through Mike Tancsa).
2001-05-13 00:03:39 +00:00
/*
* Tell the upper layer(s) we support long frames.
Cosmetic: Set capability bits in a consistent way. Add a comment on why the VLAN_MTU stuff comes after ether_ifattach().
2004-05-21 20:34:04 +00:00
* Must appear after the call to ether_ifattach() because
* ether_ifattach() sets ifi_hdrlen to the default value.
Add few cosmetic style fixes, and some debug information for SCB timeouts. Add VLAN support, obtained from Pedro J. Lobo (through Mike Tancsa).
2001-05-13 00:03:39 +00:00
*/
ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
ifp->if_capabilities |= IFCAP_VLAN_MTU;
POLA dictates that VLAN_MTU be enabled by default. In particular, disabling it was likely to break configurations involving ng_vlan(4) since the latter couldn't control the parent's VLAN_MTU in the way vlan(4) did. Pointed out by: ru
2004-05-27 14:36:58 +00:00
ifp->if_capenable |= IFCAP_VLAN_MTU; /* the hw bits already set */
Add VLAN hardware tag insertion/stripping support. Tx/Rx checksum offload for VLAN frames are also supported. The VLAN hardware assistance is available only on 82550/82551 based controllers. While I'm here change the confusing name of bit1 in byte 22 of configuration block to vlan_drop_en. The bit controls whether hardware strips VLAN tagged frame or not. Special thanks to wpaul who sent valuable VLAN related information to me. Tested on: i386, sparc64
2008-12-02 02:30:12 +00:00
if ((sc->flags & FXP_FLAG_EXT_RFA) != 0) {
ifp->if_capabilities |= IFCAP_VLAN_HWTAGGING |
IFCAP_VLAN_HWCSUM;
ifp->if_capenable |= IFCAP_VLAN_HWTAGGING |
IFCAP_VLAN_HWCSUM;
}
Add few cosmetic style fixes, and some debug information for SCB timeouts. Add VLAN support, obtained from Pedro J. Lobo (through Mike Tancsa).
2001-05-13 00:03:39 +00:00
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
/*
Implemented a more sophisticated mechanism for handling transmitter interrupts which now defers them until the transmit queue if filled up with completed buffers. This has two advantages: first, it reduces the number of transmitter interrupts to just 1/120th of the rate that they occured previously, and two, running down many buffers at once has much improved cache effects.
1998-08-04 08:53:12 +00:00
* Let the system queue as many packets as we have available
* TX descriptors.
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
*/
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_MAXLEN(&ifp->if_snd, FXP_NTXCB - 1);
ifp->if_snd.ifq_drv_maxlen = FXP_NTXCB - 1;
IFQ_SET_READY(&ifp->if_snd);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
Whitespace fix.
2008-11-25 00:48:05 +00:00
/*
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
* Hook our interrupt after all initialization is complete.
Setup our interrupt only at the end of fxp_attach(), when all other allocations/initializations have been successful. I kinda doubt it will fix the recent breakage that some people are seeing, but this could have caused problems for sure.
2003-04-05 01:27:17 +00:00
*/
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies the resource allocation code significantly.
2005-09-27 09:01:11 +00:00
error = bus_setup_intr(dev, sc->fxp_res[1], INTR_TYPE_NET | INTR_MPSAFE,
o break newbus api: add a new argument of type driver_filter_t to bus_setup_intr() o add an int return code to all fast handlers o retire INTR_FAST/IH_FAST For more info: http://docs.freebsd.org/cgi/getmsg.cgi?fetch=465712+0+current/freebsd-current Reviewed by: many Approved by: re@
2007-02-23 12:19:07 +00:00
NULL, fxp_intr, sc, &sc->ih);
Setup our interrupt only at the end of fxp_attach(), when all other allocations/initializations have been successful. I kinda doubt it will fix the recent breakage that some people are seeing, but this could have caused problems for sure.
2003-04-05 01:27:17 +00:00
if (error) {
device_printf(dev, "could not setup irq\n");
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ether_ifdetach(sc->ifp);
Setup our interrupt only at the end of fxp_attach(), when all other allocations/initializations have been successful. I kinda doubt it will fix the recent breakage that some people are seeing, but this could have caused problems for sure.
2003-04-05 01:27:17 +00:00
goto fail;
}
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
/*
* Configure hardware to reject magic frames otherwise
* system will hang on recipt of magic frames.
*/
if ((sc->flags & FXP_FLAG_WOLCAP) != 0) {
FXP_LOCK(sc);
/* Clear wakeup events. */
Make sure to clear PMDR register by writing back power management events. Just reading PMDR register was not enough to have fxp(4) immuninize against received magic packets during system boot. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-12-02 02:12:34 +00:00
CSR_WRITE_1(sc, FXP_CSR_PMDR, CSR_READ_1(sc, FXP_CSR_PMDR));
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
fxp_init_body(sc);
fxp_stop(sc);
FXP_UNLOCK(sc);
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fail:
Revert the unnecessicary addition of some braces in fxp_attach(). Don't explicitly free the ifp in fxp_detach(), the call to fxp_release() takes care of it.
2005-06-10 20:42:02 +00:00
if (error)
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
fxp_release(sc);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
return (error);
}
/*
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
* Release all resources. The softc lock should not be held and the
* interrupt should already be torn down.
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
*/
static void
fxp_release(struct fxp_softc *sc)
{
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
struct fxp_rx *rxp;
struct fxp_tx *txp;
int i;
Abstract the locking in fxp(4) a bit more by using macros for mtx_assert() and mtx_owned(), as it is done in other places, for instance proc locking.
2004-06-02 22:59:57 +00:00
FXP_LOCK_ASSERT(sc, MA_NOTOWNED);
Change a if (...) panic() to a KASSERT().
2004-05-24 18:31:56 +00:00
KASSERT(sc->ih == NULL,
("fxp_release() called with intr handle still active"));
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
if (sc->miibus)
device_delete_child(sc->dev, sc->miibus);
bus_generic_detach(sc->dev);
ifmedia_removeall(&sc->sc_media);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (sc->fxp_desc.cbl_list) {
bus_dmamap_unload(sc->cbl_tag, sc->cbl_map);
bus_dmamem_free(sc->cbl_tag, sc->fxp_desc.cbl_list,
sc->cbl_map);
}
if (sc->fxp_stats) {
bus_dmamap_unload(sc->fxp_stag, sc->fxp_smap);
bus_dmamem_free(sc->fxp_stag, sc->fxp_stats, sc->fxp_smap);
}
if (sc->mcsp) {
bus_dmamap_unload(sc->mcs_tag, sc->mcs_map);
bus_dmamem_free(sc->mcs_tag, sc->mcsp, sc->mcs_map);
}
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies the resource allocation code significantly.
2005-09-27 09:01:11 +00:00
bus_release_resources(sc->dev, sc->fxp_spec, sc->fxp_res);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
if (sc->fxp_rxmtag) {
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
for (i = 0; i < FXP_NRFABUFS; i++) {
rxp = &sc->fxp_desc.rx_list[i];
if (rxp->rx_mbuf != NULL) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_rxmtag, rxp->rx_map,
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
BUS_DMASYNC_POSTREAD);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_unload(sc->fxp_rxmtag, rxp->rx_map);
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
m_freem(rxp->rx_mbuf);
}
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_destroy(sc->fxp_rxmtag, rxp->rx_map);
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
}
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_destroy(sc->fxp_rxmtag, sc->spare_map);
bus_dma_tag_destroy(sc->fxp_rxmtag);
}
if (sc->fxp_txmtag) {
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
for (i = 0; i < FXP_NTXCB; i++) {
txp = &sc->fxp_desc.tx_list[i];
if (txp->tx_mbuf != NULL) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_txmtag, txp->tx_map,
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
BUS_DMASYNC_POSTWRITE);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_unload(sc->fxp_txmtag, txp->tx_map);
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
m_freem(txp->tx_mbuf);
}
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_destroy(sc->fxp_txmtag, txp->tx_map);
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
}
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dma_tag_destroy(sc->fxp_txmtag);
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
}
Unload and destroy the TX DMA maps before destroying the DMA tag they're attached to, not after. Spotted by: Coverity via sam
2005-03-16 16:39:04 +00:00
if (sc->fxp_stag)
bus_dma_tag_destroy(sc->fxp_stag);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (sc->cbl_tag)
bus_dma_tag_destroy(sc->cbl_tag);
if (sc->mcs_tag)
bus_dma_tag_destroy(sc->mcs_tag);
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 (sc->ifp)
if_free(sc->ifp);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
Add locking to make able to run without the Giant lock being held. This is enabling as all entries are still called with Giant being held. Maintaining compatability with NetBSD makes what should be very simple kinda ugly. Reviewed by: Jason Evans
2000-09-17 13:26:25 +00:00
mtx_destroy(&sc->sc_mtx);
Bring the 'new-bus' to the i386. This extensively changes the way the i386 platform boots, it is no longer ISA-centric, and is fully dynamic. Most old drivers compile and run without modification via 'compatability shims' to enable a smoother transition. eisa, isapnp and pccard* are not yet using the new resource manager. Once fully converted, all drivers will be loadable, including PCI and ISA. (Some other changes appear to have snuck in, including a port of Soren's ATA driver to the Alpha. Soren, back this out if you need to.) This is a checkpoint of work-in-progress, but is quite functional. The bulk of the work was done over the last few years by Doug Rabson and Garrett Wollman. Approved by: core
1999-04-16 21:22:55 +00:00
}
/*
* Detach interface.
*/
static int
fxp_detach(device_t dev)
{
struct fxp_softc *sc = device_get_softc(dev);
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
Whitespace fix.
2008-11-25 00:48:05 +00:00
if (sc->ifp->if_capenable & IFCAP_POLLING)
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
ether_poll_deregister(sc->ifp);
#endif
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_LOCK(sc);
- Call ether_ifdetach() before shutting down the hardware. - Remove a useless device_is_alive() check. - Disable interrupts if bus_child_present() so that this check is more useful. This fixes the hangs I was seeing when unloading the fxp driver. Suggestions from: hsu, njl
2003-04-11 12:52:52 +00:00
/*
Fix 5 bugs: 1) always call fxp_stop in fxp_detach. Since we don't read from the card, there's no need to carefully look at things with bus_child_present. 2) Call FXP_UNLOCK() before calling bus_teardown_intr to avoid a possible deadlock reported by jhb. 3) add gone to the softc. Set it to true in detach. 4) Return immediately if gone is true in fxp_ioctl 5) Return immediately if gone is true in fxp_intr
2003-04-29 05:45:09 +00:00
* Stop DMA and drop transmit queue, but disable interrupts first.
- Call ether_ifdetach() before shutting down the hardware. - Remove a useless device_is_alive() check. - Disable interrupts if bus_child_present() so that this check is more useful. This fixes the hangs I was seeing when unloading the fxp driver. Suggestions from: hsu, njl
2003-04-11 12:52:52 +00:00
*/
Fix 5 bugs: 1) always call fxp_stop in fxp_detach. Since we don't read from the card, there's no need to carefully look at things with bus_child_present. 2) Call FXP_UNLOCK() before calling bus_teardown_intr to avoid a possible deadlock reported by jhb. 3) add gone to the softc. Set it to true in detach. 4) Return immediately if gone is true in fxp_ioctl 5) Return immediately if gone is true in fxp_intr
2003-04-29 05:45:09 +00:00
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, FXP_SCB_INTR_DISABLE);
fxp_stop(sc);
FXP_UNLOCK(sc);
Add callout_drain()'s to foo_detach() after calling foo_stop() to make sure that if softclock is running on another CPU and is blocked on our driver lock, we will wait until it has acquired the lock, seen that it was cancelled, dropped the lock, and awakened us so that we can safely destroy the mutex. MFC after: 3 days
2005-08-17 17:44:32 +00:00
callout_drain(&sc->stat_ch);
Bring the 'new-bus' to the i386. This extensively changes the way the i386 platform boots, it is no longer ISA-centric, and is fully dynamic. Most old drivers compile and run without modification via 'compatability shims' to enable a smoother transition. eisa, isapnp and pccard* are not yet using the new resource manager. Once fully converted, all drivers will be loadable, including PCI and ISA. (Some other changes appear to have snuck in, including a port of Soren's ATA driver to the Alpha. Soren, back this out if you need to.) This is a checkpoint of work-in-progress, but is quite functional. The bulk of the work was done over the last few years by Doug Rabson and Garrett Wollman. Approved by: core
1999-04-16 21:22:55 +00:00
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
/*
* Close down routes etc.
*/
ether_ifdetach(sc->ifp);
- Call ether_ifdetach() before shutting down the hardware. - Remove a useless device_is_alive() check. - Disable interrupts if bus_child_present() so that this check is more useful. This fixes the hangs I was seeing when unloading the fxp driver. Suggestions from: hsu, njl
2003-04-11 12:52:52 +00:00
/*
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
* Unhook interrupt before dropping lock. This is to prevent
* races with fxp_intr().
- Call ether_ifdetach() before shutting down the hardware. - Remove a useless device_is_alive() check. - Disable interrupts if bus_child_present() so that this check is more useful. This fixes the hangs I was seeing when unloading the fxp driver. Suggestions from: hsu, njl
2003-04-11 12:52:52 +00:00
*/
Convert fxp(4) to use the new bus_alloc_resources() API, it simplifies the resource allocation code significantly.
2005-09-27 09:01:11 +00:00
bus_teardown_intr(sc->dev, sc->fxp_res[1], sc->ih);
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
sc->ih = NULL;
- Call ether_ifdetach() before shutting down the hardware. - Remove a useless device_is_alive() check. - Disable interrupts if bus_child_present() so that this check is more useful. This fixes the hangs I was seeing when unloading the fxp driver. Suggestions from: hsu, njl
2003-04-11 12:52:52 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/* Release our allocated resources. */
fxp_release(sc);
return (0);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
}
/*
* Device shutdown routine. Called at system shutdown after sync. The
* main purpose of this routine is to shut off receiver DMA so that
* kernel memory doesn't get clobbered during warmboot.
*/
Bring the 'new-bus' to the i386. This extensively changes the way the i386 platform boots, it is no longer ISA-centric, and is fully dynamic. Most old drivers compile and run without modification via 'compatability shims' to enable a smoother transition. eisa, isapnp and pccard* are not yet using the new resource manager. Once fully converted, all drivers will be loadable, including PCI and ISA. (Some other changes appear to have snuck in, including a port of Soren's ATA driver to the Alpha. Soren, back this out if you need to.) This is a checkpoint of work-in-progress, but is quite functional. The bulk of the work was done over the last few years by Doug Rabson and Garrett Wollman. Approved by: core
1999-04-16 21:22:55 +00:00
static int
fxp_shutdown(device_t dev)
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
{
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
Bring the 'new-bus' to the i386. This extensively changes the way the i386 platform boots, it is no longer ISA-centric, and is fully dynamic. Most old drivers compile and run without modification via 'compatability shims' to enable a smoother transition. eisa, isapnp and pccard* are not yet using the new resource manager. Once fully converted, all drivers will be loadable, including PCI and ISA. (Some other changes appear to have snuck in, including a port of Soren's ATA driver to the Alpha. Soren, back this out if you need to.) This is a checkpoint of work-in-progress, but is quite functional. The bulk of the work was done over the last few years by Doug Rabson and Garrett Wollman. Approved by: core
1999-04-16 21:22:55 +00:00
/*
* Make sure that DMA is disabled prior to reboot. Not doing
* do could allow DMA to corrupt kernel memory during the
* reboot before the driver initializes.
*/
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
return (fxp_suspend(dev));
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
}
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
/*
* Device suspend routine. Stop the interface and save some PCI
* settings in case the BIOS doesn't restore them properly on
* resume.
*/
static int
fxp_suspend(device_t dev)
{
struct fxp_softc *sc = device_get_softc(dev);
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
struct ifnet *ifp;
int pmc;
uint16_t pmstat;
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_LOCK(sc);
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
ifp = sc->ifp;
if (pci_find_extcap(sc->dev, PCIY_PMG, &pmc) == 0) {
pmstat = pci_read_config(sc->dev, pmc + PCIR_POWER_STATUS, 2);
pmstat &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
if ((ifp->if_capenable & IFCAP_WOL_MAGIC) != 0) {
/* Request PME. */
pmstat |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
sc->flags |= FXP_FLAG_WOL;
/* Reconfigure hardware to accept magic frames. */
fxp_init_body(sc);
}
pci_write_config(sc->dev, pmc + PCIR_POWER_STATUS, pmstat, 2);
}
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
fxp_stop(sc);
Whitespace fix.
2008-11-25 00:48:05 +00:00
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
sc->suspended = 1;
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_UNLOCK(sc);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
return (0);
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
}
/*
Correct comment
2005-06-06 03:22:51 +00:00
* Device resume routine. re-enable busmastering, and restart the interface if
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
* appropriate.
*/
static int
fxp_resume(device_t dev)
{
struct fxp_softc *sc = device_get_softc(dev);
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
struct ifnet *ifp = sc->ifp;
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
int pmc;
uint16_t pmstat;
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_LOCK(sc);
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
if (pci_find_extcap(sc->dev, PCIY_PMG, &pmc) == 0) {
sc->flags &= ~FXP_FLAG_WOL;
pmstat = pci_read_config(sc->dev, pmc + PCIR_POWER_STATUS, 2);
/* Disable PME and clear PME status. */
pmstat &= ~PCIM_PSTAT_PMEENABLE;
pci_write_config(sc->dev, pmc + PCIR_POWER_STATUS, pmstat, 2);
Make sure to clear PMDR register by writing back power management events. Just reading PMDR register was not enough to have fxp(4) immuninize against received magic packets during system boot. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-12-02 02:12:34 +00:00
if ((sc->flags & FXP_FLAG_WOLCAP) != 0)
CSR_WRITE_1(sc, FXP_CSR_PMDR,
CSR_READ_1(sc, FXP_CSR_PMDR));
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
}
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SELECTIVE_RESET);
DELAY(10);
/* reinitialize interface if necessary */
if (ifp->if_flags & IFF_UP)
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
fxp_init_body(sc);
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
sc->suspended = 0;
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_UNLOCK(sc);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
return (0);
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
}
Whitespace fix.
2008-11-25 00:48:05 +00:00
static void
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
fxp_eeprom_shiftin(struct fxp_softc *sc, int data, int length)
{
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint16_t reg;
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
int x;
/*
* Shift in data.
*/
for (x = 1 << (length - 1); x; x >>= 1) {
if (data & x)
reg = FXP_EEPROM_EECS | FXP_EEPROM_EEDI;
else
reg = FXP_EEPROM_EECS;
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg);
DELAY(1);
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg | FXP_EEPROM_EESK);
DELAY(1);
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg);
DELAY(1);
}
}
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
/*
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
* Read from the serial EEPROM. Basically, you manually shift in
* the read opcode (one bit at a time) and then shift in the address,
* and then you shift out the data (all of this one bit at a time).
* The word size is 16 bits, so you have to provide the address for
* every 16 bits of data.
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
*/
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
static uint16_t
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_eeprom_getword(struct fxp_softc *sc, int offset, int autosize)
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
{
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint16_t reg, data;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
int x;
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
/*
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
* Shift in read opcode.
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
*/
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
fxp_eeprom_shiftin(sc, FXP_EEPROM_OPC_READ, 3);
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
/*
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
* Shift in address.
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
*/
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
data = 0;
for (x = 1 << (sc->eeprom_size - 1); x; x >>= 1) {
if (offset & x)
reg = FXP_EEPROM_EECS | FXP_EEPROM_EEDI;
else
reg = FXP_EEPROM_EECS;
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg);
DELAY(1);
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg | FXP_EEPROM_EESK);
DELAY(1);
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg);
DELAY(1);
reg = CSR_READ_2(sc, FXP_CSR_EEPROMCONTROL) & FXP_EEPROM_EEDO;
data++;
if (autosize && reg == 0) {
sc->eeprom_size = data;
break;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
}
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
/*
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
* Shift out data.
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
*/
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
data = 0;
reg = FXP_EEPROM_EECS;
for (x = 1 << 15; x; x >>= 1) {
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg | FXP_EEPROM_EESK);
DELAY(1);
if (CSR_READ_2(sc, FXP_CSR_EEPROMCONTROL) & FXP_EEPROM_EEDO)
data |= x;
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, reg);
DELAY(1);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
DELAY(1);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
return (data);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
static void
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
fxp_eeprom_putword(struct fxp_softc *sc, int offset, uint16_t data)
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
{
int i;
/*
* Erase/write enable.
*/
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
fxp_eeprom_shiftin(sc, 0x4, 3);
fxp_eeprom_shiftin(sc, 0x03 << (sc->eeprom_size - 2), sc->eeprom_size);
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
DELAY(1);
/*
* Shift in write opcode, address, data.
*/
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
fxp_eeprom_shiftin(sc, FXP_EEPROM_OPC_WRITE, 3);
fxp_eeprom_shiftin(sc, offset, sc->eeprom_size);
fxp_eeprom_shiftin(sc, data, 16);
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
DELAY(1);
/*
* Wait for EEPROM to finish up.
*/
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
DELAY(1);
for (i = 0; i < 1000; i++) {
if (CSR_READ_2(sc, FXP_CSR_EEPROMCONTROL) & FXP_EEPROM_EEDO)
break;
DELAY(50);
}
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
DELAY(1);
/*
* Erase/write disable.
*/
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, FXP_EEPROM_EECS);
fxp_eeprom_shiftin(sc, 0x4, 3);
fxp_eeprom_shiftin(sc, 0, sc->eeprom_size);
CSR_WRITE_2(sc, FXP_CSR_EEPROMCONTROL, 0);
DELAY(1);
}
Added support for cards and on-motherboard NICs that use an SEEPROM address size that is different than the standard 6bits. This fixes support for the Compaq NC3121 card, certain newer Intel Pro/100+ cards, and should also fix integrated NICs on SuperMicro and Compaq motherboards. The auto-sizing algorithm was taken from NetBSD (thanks!), which I think got it from Linux originally. Thanks also to Andrew Sparrow <spadger@best.com> and Joe Moore <jomor@ahpcns.com> for supplying me with unworking Compaq and Intel cards to develop and test the fixes with.
2000-03-28 04:41:42 +00:00
/*
* From NetBSD:
*
* Figure out EEPROM size.
*
* 559's can have either 64-word or 256-word EEPROMs, the 558
* datasheet only talks about 64-word EEPROMs, and the 557 datasheet
* talks about the existance of 16 to 256 word EEPROMs.
*
* The only known sizes are 64 and 256, where the 256 version is used
* by CardBus cards to store CIS information.
*
* The address is shifted in msb-to-lsb, and after the last
* address-bit the EEPROM is supposed to output a `dummy zero' bit,
* after which follows the actual data. We try to detect this zero, by
* probing the data-out bit in the EEPROM control register just after
* having shifted in a bit. If the bit is zero, we assume we've
* shifted enough address bits. The data-out should be tri-state,
* before this, which should translate to a logical one.
*/
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_autosize_eeprom(struct fxp_softc *sc)
Added support for cards and on-motherboard NICs that use an SEEPROM address size that is different than the standard 6bits. This fixes support for the Compaq NC3121 card, certain newer Intel Pro/100+ cards, and should also fix integrated NICs on SuperMicro and Compaq motherboards. The auto-sizing algorithm was taken from NetBSD (thanks!), which I think got it from Linux originally. Thanks also to Andrew Sparrow <spadger@best.com> and Joe Moore <jomor@ahpcns.com> for supplying me with unworking Compaq and Intel cards to develop and test the fixes with.
2000-03-28 04:41:42 +00:00
{
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
/* guess maximum size of 256 words */
sc->eeprom_size = 8;
/* autosize */
(void) fxp_eeprom_getword(sc, 0, 1);
Added support for cards and on-motherboard NICs that use an SEEPROM address size that is different than the standard 6bits. This fixes support for the Compaq NC3121 card, certain newer Intel Pro/100+ cards, and should also fix integrated NICs on SuperMicro and Compaq motherboards. The auto-sizing algorithm was taken from NetBSD (thanks!), which I think got it from Linux originally. Thanks also to Andrew Sparrow <spadger@best.com> and Joe Moore <jomor@ahpcns.com> for supplying me with unworking Compaq and Intel cards to develop and test the fixes with.
2000-03-28 04:41:42 +00:00
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_read_eeprom(struct fxp_softc *sc, u_short *data, int offset, int words)
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
{
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
int i;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
for (i = 0; i < words; i++)
data[i] = fxp_eeprom_getword(sc, offset + i, 0);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Systems based on the ICH2/ICH2-M chip from Intel have a defect where the chip can cause a PCI protocol violation in under certain scenarios. The workaround is to rewrite the EEPROM to disable Dynamic Standby Mode. Once the EEPROM is rewritten, the system needs to be rebooted in order to pick up the new settings. This has been tested on several ICH2/ICH2-M systems, found in 815E based boards, and usually identified by the presence of the 82562 ET/EM PHY. Thanks to: Mike Tansca, Paul Saab for samples of the problematic boards.
2001-08-27 16:07:12 +00:00
static void
fxp_write_eeprom(struct fxp_softc *sc, u_short *data, int offset, int words)
{
int i;
for (i = 0; i < words; i++)
fxp_eeprom_putword(sc, offset + i, data[i]);
}
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
* Grab the softc lock and call the real fxp_start_body() routine
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
*/
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_start(struct ifnet *ifp)
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
{
struct fxp_softc *sc = ifp->if_softc;
FXP_LOCK(sc);
fxp_start_body(ifp);
FXP_UNLOCK(sc);
}
/*
Whitespace fix.
2008-11-25 00:48:05 +00:00
* Start packet transmission on the interface.
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
* This routine must be called with the softc lock held, and is an
* internal entry point only.
*/
static void
fxp_start_body(struct ifnet *ifp)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Clean up Ethernet drivers: - fill in and use ifp->if_softc - use if_bpf rather than private cookie variables - change bpf interface to take advantage of this - call ether_ifattach() directly from Ethernet drivers - delete kludge in if_attach() that did this indirectly
1996-02-06 18:51:28 +00:00
struct fxp_softc *sc = ifp->if_softc;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
struct mbuf *mb_head;
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
int txqueued;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Abstract the locking in fxp(4) a bit more by using macros for mtx_assert() and mtx_owned(), as it is done in other places, for instance proc locking.
2004-06-02 22:59:57 +00:00
FXP_LOCK_ASSERT(sc, MA_OWNED);
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
Always check fxp(4) is running, see if it can accept frames from upper stack in fxp_start_body(). fxp(4) drops driver lock in Rx path so check the fxp(4) is still running after reacquiring driver lock in Rx path. Also don't invoke fxp_intr_body if fxp(4) is not running. With this change there is no need to set suspend bit in device attach phase.
2009-06-21 06:46:32 +00:00
if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
IFF_DRV_RUNNING)
return;
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
if (sc->tx_queued > FXP_NTXCB_HIWAT)
fxp_txeof(sc);
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
/*
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
* We're finished if there is nothing more to add to the list or if
* we're all filled up with buffers to transmit.
Implemented a more sophisticated mechanism for handling transmitter interrupts which now defers them until the transmit queue if filled up with completed buffers. This has two advantages: first, it reduces the number of transmitter interrupts to just 1/120th of the rate that they occured previously, and two, running down many buffers at once has much improved cache effects.
1998-08-04 08:53:12 +00:00
* NOTE: One TxCB is reserved to guarantee that fxp_mc_setup() can add
* a NOP command when needed.
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
*/
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
txqueued = 0;
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
while (!IFQ_DRV_IS_EMPTY(&ifp->if_snd) &&
sc->tx_queued < FXP_NTXCB - 1) {
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
* Grab a packet to transmit.
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
*/
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_DEQUEUE(&ifp->if_snd, mb_head);
if (mb_head == NULL)
break;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
if (fxp_encap(sc, &mb_head)) {
if (mb_head == NULL)
break;
IFQ_DRV_PREPEND(&ifp->if_snd, mb_head);
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
}
txqueued++;
/*
* Pass packet to bpf if there is a listener.
*/
BPF_MTAP(ifp, mb_head);
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
}
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
/*
* We're finished. If we added to the list, issue a RESUME to get DMA
* going again if suspended.
*/
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
if (txqueued > 0) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
fxp_scb_wait(sc);
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_RESUME);
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
/*
* Set a 5 second timer just in case we don't hear
* from the card again.
*/
sc->watchdog_timer = 5;
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
}
}
Always set the hardware parse bit in the IPCB structure when this structure, which is new to the 82550 and 82551, is used to transmit a packet. This appears to fix the packet truncation problem that was observed when using 82550-based fxp cards to transmit ICMP or fragmented UDP packets of certain lengths which only had one to three bytes in the second and final mbuf of the packet. This matches a note in the "Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual", which says that the hardware parse bit should be set when sending these types of packets. There have also been unconfirmed reports of similar problems when transmitting TCP packets, which should not be affected by the above mentioned change because the hardware parse bit was already being set if the stack requested hardware checksumming of the packet. If the problem remains, the use of the IPCB structure can be disabled to cause the driver to fall back to using the older 82559 interface with 82550-based cards by setting hint.fxp.UNIT_NUMBER.ipcbxmit_disable to a non-zero value at boot time, or using kenv to set this variable before using kldload to load the fxp driver. Approved by: re (jhb)
2003-05-25 05:04:26 +00:00
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
static int
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
fxp_encap(struct fxp_softc *sc, struct mbuf **m_head)
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
{
struct ifnet *ifp;
struct mbuf *m;
struct fxp_tx *txp;
struct fxp_cb_tx *cbp;
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
struct tcphdr *tcp;
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
bus_dma_segment_t segs[FXP_NTXSEG];
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
int error, i, nseg, tcp_payload;
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
FXP_LOCK_ASSERT(sc, MA_OWNED);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->ifp;
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
tcp_payload = 0;
tcp = NULL;
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
/*
* Get pointer to next available tx desc.
*/
txp = sc->fxp_desc.tx_last->tx_next;
/*
* A note in Appendix B of the Intel 8255x 10/100 Mbps
* Ethernet Controller Family Open Source Software
* Developer Manual says:
* Using software parsing is only allowed with legal
* TCP/IP or UDP/IP packets.
* ...
* For all other datagrams, hardware parsing must
* be used.
* Software parsing appears to truncate ICMP and
* fragmented UDP packets that contain one to three
* bytes in the second (and final) mbuf of the packet.
*/
if (sc->flags & FXP_FLAG_EXT_RFA)
txp->tx_cb->ipcb_ip_activation_high =
FXP_IPCB_HARDWAREPARSING_ENABLE;
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
m = *m_head;
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
if (m->m_pkthdr.csum_flags & CSUM_TSO) {
/*
* 82550/82551 requires ethernet/IP/TCP headers must be
* contained in the first active transmit buffer.
*/
struct ether_header *eh;
struct ip *ip;
uint32_t ip_off, poff;
if (M_WRITABLE(*m_head) == 0) {
/* Get a writable copy. */
m = m_dup(*m_head, M_DONTWAIT);
m_freem(*m_head);
if (m == NULL) {
*m_head = NULL;
return (ENOBUFS);
}
*m_head = m;
}
ip_off = sizeof(struct ether_header);
m = m_pullup(*m_head, ip_off);
if (m == NULL) {
*m_head = NULL;
return (ENOBUFS);
}
eh = mtod(m, struct ether_header *);
/* Check the existence of VLAN tag. */
if (eh->ether_type == htons(ETHERTYPE_VLAN)) {
ip_off = sizeof(struct ether_vlan_header);
m = m_pullup(m, ip_off);
if (m == NULL) {
*m_head = NULL;
return (ENOBUFS);
}
}
m = m_pullup(m, ip_off + sizeof(struct ip));
if (m == NULL) {
*m_head = NULL;
return (ENOBUFS);
}
ip = (struct ip *)(mtod(m, char *) + ip_off);
poff = ip_off + (ip->ip_hl << 2);
m = m_pullup(m, poff + sizeof(struct tcphdr));
if (m == NULL) {
*m_head = NULL;
return (ENOBUFS);
}
tcp = (struct tcphdr *)(mtod(m, char *) + poff);
m = m_pullup(m, poff + sizeof(struct tcphdr) + tcp->th_off);
if (m == NULL) {
*m_head = NULL;
return (ENOBUFS);
}
/*
* Since 82550/82551 doesn't modify IP length and pseudo
* checksum in the first frame driver should compute it.
*/
ip->ip_sum = 0;
Don't blindly set IP packet length from interface MTU in TSO case. Remote host can advertise smaller MSS than that of sender so upper stack might have adjusted the MSS which in turn generates IP packets that are less size than that of interface MTU. Reported by: Bjoern Koenig ( bkoenig <> alpha-tierchen dot de ) Tested by: Bjoern Koenig ( bkoenig <> alpha-tierchen dot de ) MFC after: 3 days
2009-04-13 00:32:34 +00:00
ip->ip_len = htons(m->m_pkthdr.tso_segsz + (ip->ip_hl << 2) +
(tcp->th_off << 2));
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
tcp->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr,
htons(IPPROTO_TCP + (tcp->th_off << 2) +
m->m_pkthdr.tso_segsz));
/* Compute total TCP payload. */
tcp_payload = m->m_pkthdr.len - ip_off - (ip->ip_hl << 2);
tcp_payload -= tcp->th_off << 2;
*m_head = m;
With r206844, CSUM_TCP is also set for CSUM_TSO case. Modify drivers to take into account for the change. Basically CSUM_TSO should be checked before checking CSUM_TCP.
2010-04-19 22:10:40 +00:00
} else if (m->m_pkthdr.csum_flags & FXP_CSUM_FEATURES) {
/*
* Deal with TCP/IP checksum offload. Note that
* in order for TCP checksum offload to work,
* the pseudo header checksum must have already
* been computed and stored in the checksum field
* in the TCP header. The stack should have
* already done this for us.
*/
txp->tx_cb->ipcb_ip_schedule = FXP_IPCB_TCPUDP_CHECKSUM_ENABLE;
if (m->m_pkthdr.csum_flags & CSUM_TCP)
txp->tx_cb->ipcb_ip_schedule |= FXP_IPCB_TCP_PACKET;
#ifdef FXP_IP_CSUM_WAR
/*
* XXX The 82550 chip appears to have trouble
* dealing with IP header checksums in very small
* datagrams, namely fragments from 1 to 3 bytes
* in size. For example, say you want to transmit
* a UDP packet of 1473 bytes. The packet will be
* fragmented over two IP datagrams, the latter
* containing only one byte of data. The 82550 will
* botch the header checksum on the 1-byte fragment.
* As long as the datagram contains 4 or more bytes
* of data, you're ok.
*
* The following code attempts to work around this
* problem: if the datagram is less than 38 bytes
* in size (14 bytes ether header, 20 bytes IP header,
* plus 4 bytes of data), we punt and compute the IP
* header checksum by hand. This workaround doesn't
* work very well, however, since it can be fooled
* by things like VLAN tags and IP options that make
* the header sizes/offsets vary.
*/
if (m->m_pkthdr.csum_flags & CSUM_IP) {
if (m->m_pkthdr.len < 38) {
struct ip *ip;
m->m_data += ETHER_HDR_LEN;
ip = mtod(m, struct ip *);
ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
m->m_data -= ETHER_HDR_LEN;
m->m_pkthdr.csum_flags &= ~CSUM_IP;
} else {
txp->tx_cb->ipcb_ip_activation_high =
FXP_IPCB_HARDWAREPARSING_ENABLE;
txp->tx_cb->ipcb_ip_schedule |=
FXP_IPCB_IP_CHECKSUM_ENABLE;
}
}
#endif
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
}
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
error = bus_dmamap_load_mbuf_sg(sc->fxp_txmtag, txp->tx_map, *m_head,
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
segs, &nseg, 0);
if (error == EFBIG) {
m = m_collapse(*m_head, M_DONTWAIT, sc->maxtxseg);
if (m == NULL) {
m_freem(*m_head);
*m_head = NULL;
return (ENOMEM);
Increased the number of Tx segments from 13 to 29 to reduce the need to recopy to near zero. Wrote the necessary code to recopy the mbuf chain into another buffer if there are too many mbufs in the chain.
1995-12-05 11:49:55 +00:00
}
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
*m_head = m;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
error = bus_dmamap_load_mbuf_sg(sc->fxp_txmtag, txp->tx_map,
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
*m_head, segs, &nseg, 0);
if (error != 0) {
m_freem(*m_head);
*m_head = NULL;
return (ENOMEM);
}
} else if (error != 0)
return (error);
if (nseg == 0) {
m_freem(*m_head);
*m_head = NULL;
return (EIO);
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
}
Increased the number of Tx segments from 13 to 29 to reduce the need to recopy to near zero. Wrote the necessary code to recopy the mbuf chain into another buffer if there are too many mbufs in the chain.
1995-12-05 11:49:55 +00:00
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
KASSERT(nseg <= sc->maxtxseg, ("too many DMA segments"));
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_txmtag, txp->tx_map, BUS_DMASYNC_PREWRITE);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
cbp = txp->tx_cb;
for (i = 0; i < nseg; i++) {
/*
* If this is an 82550/82551, then we're using extended
* TxCBs _and_ we're using checksum offload. This means
* that the TxCB is really an IPCB. One major difference
* between the two is that with plain extended TxCBs,
* the bottom half of the TxCB contains two entries from
* the TBD array, whereas IPCBs contain just one entry:
* one entry (8 bytes) has been sacrificed for the TCP/IP
* checksum offload control bits. So to make things work
* right, we have to start filling in the TBD array
* starting from a different place depending on whether
* the chip is an 82550/82551 or not.
*/
if (sc->flags & FXP_FLAG_EXT_RFA) {
It seems that we don't need to reserve a TBD to set total TCP payload length in TSO case. Leaving unused TBD also seem to cause SCB timeouts under certain conditions when TSO/non-TSO traffics are active at the same time.
2008-12-18 01:36:46 +00:00
cbp->tbd[i + 1].tb_addr = htole32(segs[i].ds_addr);
cbp->tbd[i + 1].tb_size = htole32(segs[i].ds_len);
Implemented a more sophisticated mechanism for handling transmitter interrupts which now defers them until the transmit queue if filled up with completed buffers. This has two advantages: first, it reduces the number of transmitter interrupts to just 1/120th of the rate that they occured previously, and two, running down many buffers at once has much improved cache effects.
1998-08-04 08:53:12 +00:00
} else {
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
cbp->tbd[i].tb_addr = htole32(segs[i].ds_addr);
cbp->tbd[i].tb_size = htole32(segs[i].ds_len);
Implemented a more sophisticated mechanism for handling transmitter interrupts which now defers them until the transmit queue if filled up with completed buffers. This has two advantages: first, it reduces the number of transmitter interrupts to just 1/120th of the rate that they occured previously, and two, running down many buffers at once has much improved cache effects.
1998-08-04 08:53:12 +00:00
}
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
}
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
if (sc->flags & FXP_FLAG_EXT_RFA) {
/* Configure dynamic TBD for 82550/82551. */
cbp->tbd_number = 0xFF;
It seems that we don't need to reserve a TBD to set total TCP payload length in TSO case. Leaving unused TBD also seem to cause SCB timeouts under certain conditions when TSO/non-TSO traffics are active at the same time.
2008-12-18 01:36:46 +00:00
cbp->tbd[nseg].tb_size |= htole32(0x8000);
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
} else
cbp->tbd_number = nseg;
/* Configure TSO. */
if (m->m_pkthdr.csum_flags & CSUM_TSO) {
cbp->tbd[-1].tb_size = htole32(m->m_pkthdr.tso_segsz << 16);
It seems that we don't need to reserve a TBD to set total TCP payload length in TSO case. Leaving unused TBD also seem to cause SCB timeouts under certain conditions when TSO/non-TSO traffics are active at the same time.
2008-12-18 01:36:46 +00:00
cbp->tbd[1].tb_size |= htole32(tcp_payload << 16);
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
cbp->ipcb_ip_schedule |= FXP_IPCB_LARGESEND_ENABLE |
FXP_IPCB_IP_CHECKSUM_ENABLE |
FXP_IPCB_TCP_PACKET |
FXP_IPCB_TCPUDP_CHECKSUM_ENABLE;
}
Add VLAN hardware tag insertion/stripping support. Tx/Rx checksum offload for VLAN frames are also supported. The VLAN hardware assistance is available only on 82550/82551 based controllers. While I'm here change the confusing name of bit1 in byte 22 of configuration block to vlan_drop_en. The bit controls whether hardware strips VLAN tagged frame or not. Special thanks to wpaul who sent valuable VLAN related information to me. Tested on: i386, sparc64
2008-12-02 02:30:12 +00:00
/* Configure VLAN hardware tag insertion. */
if ((m->m_flags & M_VLANTAG) != 0) {
cbp->ipcb_vlan_id = htons(m->m_pkthdr.ether_vtag);
txp->tx_cb->ipcb_ip_activation_high |=
FXP_IPCB_INSERTVLAN_ENABLE;
}
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
txp->tx_mbuf = m;
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
txp->tx_cb->cb_status = 0;
txp->tx_cb->byte_count = 0;
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
if (sc->tx_queued != FXP_CXINT_THRESH - 1)
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
txp->tx_cb->cb_command =
htole16(sc->tx_cmd | FXP_CB_COMMAND_SF |
FXP_CB_COMMAND_S);
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
else
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
txp->tx_cb->cb_command =
htole16(sc->tx_cmd | FXP_CB_COMMAND_SF |
FXP_CB_COMMAND_S | FXP_CB_COMMAND_I);
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
if ((m->m_pkthdr.csum_flags & CSUM_TSO) == 0)
txp->tx_cb->tx_threshold = tx_threshold;
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
/*
* Advance the end of list forward.
*/
Fix an alpha-only race which causes the transmit side of the chip to lock up under moderate to heavy load. The status & command fields share a 32-bit longword. The programming API of the eepro apparently requires that you update the command field of a transmit slot that you've already given to the card. This means the card could be updating the status field of the same longword at the same time. Since alphas can only operate on 32-bit chunks of memory, both the status & command fields are loaded from memory & operated on in registers when the following line of C is executed: sc->cbl_last->cb_command &= ~FXP_CB_COMMAND_S; The race is caused by the card DMA'ing up the status at just the wrong time -- after it has been loaded into a register & before it has been written back. The old value of the status is written back, clobbering the status the card just DMA'ed up. The fact that the card has sent this frame is missed & the transmit engine appears to hang. Luckily, as numerous people on the freebsd-alpha list pointed out, the load-locked/store-conditional instructions used by the atomic functions work with respect changes in memory due to I/O devices. We now use them to safely update the command field. Tested by: Bernd Walter <ticso@mail.cicely.de>
2000-07-19 14:33:52 +00:00
Fix the hangs people have been seeing. For some reason, the atomic operation change caused this, so backout it until I can figure out what's wrong with it.
2003-04-08 13:05:11 +00:00
#ifdef __alpha__
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
/*
* On platforms which can't access memory in 16-bit
* granularities, we must prevent the card from DMA'ing
* up the status while we update the command field.
* This could cause us to overwrite the completion status.
* XXX This is probably bogus and we're _not_ looking
* for atomicity here.
*/
atomic_clear_16(&sc->fxp_desc.tx_last->tx_cb->cb_command,
htole16(FXP_CB_COMMAND_S));
Fix the hangs people have been seeing. For some reason, the atomic operation change caused this, so backout it until I can figure out what's wrong with it.
2003-04-08 13:05:11 +00:00
#else
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
sc->fxp_desc.tx_last->tx_cb->cb_command &= htole16(~FXP_CB_COMMAND_S);
Fix the hangs people have been seeing. For some reason, the atomic operation change caused this, so backout it until I can figure out what's wrong with it.
2003-04-08 13:05:11 +00:00
#endif /*__alpha__*/
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
sc->fxp_desc.tx_last = txp;
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
/*
* Advance the beginning of the list forward if there are
* no other packets queued (when nothing is queued, tx_first
* sits on the last TxCB that was sent out).
*/
if (sc->tx_queued == 0)
sc->fxp_desc.tx_first = txp;
Backed out one of my "performance optimizations" as it results in sometimes not resuming the NIC as required for transmit. Thanks to Alan Cox <alc@cs.rice.edu> for noticing this. Added another performance optimization to compensate. :-) Changed crscdt to 1...strange, but this seems to be needed for some reason despite what the manual says.
1996-09-29 10:20:45 +00:00
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
sc->tx_queued++;
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
- Encapsulate the code responsible for initializing a new TX descriptor from an mbuf into the fxp_encap() function, as done in other drivers. - Don't waste time calling bus_dmamap_load_mbuf() if we know the mbuf chain is too long to fit in a TX descriptor, call m_defrag() first. - Convert fxp(4) to use bus_dmamap_load_mbuf_sg().
2005-03-07 13:20:49 +00:00
return (0);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
#ifdef DEVICE_POLLING
static poll_handler_t fxp_poll;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
fxp_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
{
struct fxp_softc *sc = ifp->if_softc;
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint8_t statack;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int rx_npkts = 0;
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_LOCK(sc);
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
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_UNLOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
}
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
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
statack = FXP_SCB_STATACK_CXTNO | FXP_SCB_STATACK_CNA |
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
FXP_SCB_STATACK_FR;
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
if (cmd == POLL_AND_CHECK_STATUS) {
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint8_t tmp;
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
tmp = CSR_READ_1(sc, FXP_CSR_SCB_STATACK);
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
if (tmp == 0xff || tmp == 0) {
FXP_UNLOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts); /* nothing to do */
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
}
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
tmp &= ~statack;
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
/* ack what we can */
if (tmp != 0)
CSR_WRITE_1(sc, FXP_CSR_SCB_STATACK, tmp);
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
statack |= tmp;
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
}
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = fxp_intr_body(sc, ifp, statack, count);
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_UNLOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
}
#endif /* DEVICE_POLLING */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
Fixed obsolete comment.
1996-11-18 02:45:46 +00:00
* Process interface interrupts.
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
*/
Removed NetBSD support, which bit-rotted long ago. Changed new SMP locking macros given the new situation.
2000-09-18 21:12:19 +00:00
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_intr(void *xsc)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
struct fxp_softc *sc = xsc;
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
struct ifnet *ifp = sc->ifp;
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint8_t statack;
Add locking to make able to run without the Giant lock being held. This is enabling as all entries are still called with Giant being held. Maintaining compatability with NetBSD makes what should be very simple kinda ugly. Reviewed by: Jason Evans
2000-09-17 13:26:25 +00:00
Get rid of the redundant 'gone' field, and overload suspend instead. Check for suspend before the device polling, rather than after it. Check to see if the current thread owns the lock in ioctl and return EBUSY if it does. This advances the locking to the point that I can eject my fxp card 10 times in a row, but I agree with Jeff Hsu that we need to get the network layer locking finished before chasing more of the races here (actually, he doesn't think this set is worth it even). There's a number of races between FXP_LOCK in detach and all other users of FXP_LOCK, and this gets back to the 'device with sleepers being forcibly detached' problem as well...
2003-04-30 01:54:38 +00:00
FXP_LOCK(sc);
if (sc->suspended) {
FXP_UNLOCK(sc);
(5) from last commit was omitted by mistake: (5) Return immediately from fxp_intr() if gone is set.
2003-04-29 05:47:14 +00:00
return;
Get rid of the redundant 'gone' field, and overload suspend instead. Check for suspend before the device polling, rather than after it. Check to see if the current thread owns the lock in ioctl and return EBUSY if it does. This advances the locking to the point that I can eject my fxp card 10 times in a row, but I agree with Jeff Hsu that we need to get the network layer locking finished before chasing more of the races here (actually, he doesn't think this set is worth it even). There's a number of races between FXP_LOCK in detach and all other users of FXP_LOCK, and this gets back to the 'device with sleepers being forcibly detached' problem as well...
2003-04-30 01:54:38 +00:00
}
(5) from last commit was omitted by mistake: (5) Return immediately from fxp_intr() if gone is set.
2003-04-29 05:47:14 +00:00
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
#ifdef DEVICE_POLLING
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
if (ifp->if_capenable & IFCAP_POLLING) {
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_UNLOCK(sc);
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
return;
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
}
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
#endif
As a minor optimization, do suspended checking more like it was originally in the PR - before the while loop.
2000-09-17 23:04:57 +00:00
while ((statack = CSR_READ_1(sc, FXP_CSR_SCB_STATACK)) != 0) {
While in the interrupt loop, check for a bogus interrupt value of 0xff. This may be returned if the underlying hardware is a pc-card which has been ejected. Reviewed by: warner
2001-06-04 22:01:44 +00:00
/*
* It should not be possible to have all bits set; the
Whitespace fix.
2008-11-25 00:48:05 +00:00
* FXP_SCB_INTR_SWI bit always returns 0 on a read. If
While in the interrupt loop, check for a bogus interrupt value of 0xff. This may be returned if the underlying hardware is a pc-card which has been ejected. Reviewed by: warner
2001-06-04 22:01:44 +00:00
* all bits are set, this may indicate that the card has
* been physically ejected, so ignore it.
Whitespace fix.
2008-11-25 00:48:05 +00:00
*/
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
if (statack == 0xff) {
FXP_UNLOCK(sc);
While in the interrupt loop, check for a bogus interrupt value of 0xff. This may be returned if the underlying hardware is a pc-card which has been ejected. Reviewed by: warner
2001-06-04 22:01:44 +00:00
return;
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
}
While in the interrupt loop, check for a bogus interrupt value of 0xff. This may be returned if the underlying hardware is a pc-card which has been ejected. Reviewed by: warner
2001-06-04 22:01:44 +00:00
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* First ACK all the interrupts in this pass.
*/
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
CSR_WRITE_1(sc, FXP_CSR_SCB_STATACK, statack);
Always check fxp(4) is running, see if it can accept frames from upper stack in fxp_start_body(). fxp(4) drops driver lock in Rx path so check the fxp(4) is still running after reacquiring driver lock in Rx path. Also don't invoke fxp_intr_body if fxp(4) is not running. With this change there is no need to set suspend bit in device attach phase.
2009-06-21 06:46:32 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
fxp_intr_body(sc, ifp, statack, -1);
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
}
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_UNLOCK(sc);
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
}
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
static void
fxp_txeof(struct fxp_softc *sc)
{
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
struct ifnet *ifp;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
struct fxp_tx *txp;
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
ifp = sc->ifp;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
for (txp = sc->fxp_desc.tx_first; sc->tx_queued &&
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
(le16toh(txp->tx_cb->cb_status) & FXP_CB_STATUS_C) != 0;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
txp = txp->tx_next) {
if (txp->tx_mbuf != NULL) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_txmtag, txp->tx_map,
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
BUS_DMASYNC_POSTWRITE);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_unload(sc->fxp_txmtag, txp->tx_map);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
m_freem(txp->tx_mbuf);
txp->tx_mbuf = NULL;
/* clear this to reset csum offload bits */
txp->tx_cb->tbd[0].tb_addr = 0;
}
sc->tx_queued--;
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
}
sc->fxp_desc.tx_first = txp;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
if (sc->tx_queued == 0)
Move unarming watchdog timer and pending multicast configuration check to fxp_txeof(). While I'm here unarm watchdog timer only if there are no pending queued Tx requests. Previously the watchdog timer was unarmed whenever Tx interrupt is raised. This could be resulted in hiding root cause of watchdog timeouts.
2008-11-25 02:05:01 +00:00
sc->watchdog_timer = 0;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
}
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
static void
fxp_rxcsum(struct fxp_softc *sc, struct ifnet *ifp, struct mbuf *m,
uint16_t status, int pos)
{
struct ether_header *eh;
struct ip *ip;
struct udphdr *uh;
int32_t hlen, len, pktlen, temp32;
uint16_t csum, *opts;
if ((sc->flags & FXP_FLAG_82559_RXCSUM) == 0) {
if ((status & FXP_RFA_STATUS_PARSE) != 0) {
if (status & FXP_RFDX_CS_IP_CSUM_BIT_VALID)
m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
if (status & FXP_RFDX_CS_IP_CSUM_VALID)
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
if ((status & FXP_RFDX_CS_TCPUDP_CSUM_BIT_VALID) &&
(status & FXP_RFDX_CS_TCPUDP_CSUM_VALID)) {
m->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
CSUM_PSEUDO_HDR;
m->m_pkthdr.csum_data = 0xffff;
}
}
return;
}
pktlen = m->m_pkthdr.len;
if (pktlen < sizeof(struct ether_header) + sizeof(struct ip))
return;
eh = mtod(m, struct ether_header *);
if (eh->ether_type != htons(ETHERTYPE_IP))
return;
ip = (struct ip *)(eh + 1);
if (ip->ip_v != IPVERSION)
return;
hlen = ip->ip_hl << 2;
pktlen -= sizeof(struct ether_header);
if (hlen < sizeof(struct ip))
return;
if (ntohs(ip->ip_len) < hlen)
return;
if (ntohs(ip->ip_len) != pktlen)
return;
if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
return; /* can't handle fragmented packet */
switch (ip->ip_p) {
case IPPROTO_TCP:
if (pktlen < (hlen + sizeof(struct tcphdr)))
return;
break;
case IPPROTO_UDP:
if (pktlen < (hlen + sizeof(struct udphdr)))
return;
uh = (struct udphdr *)((caddr_t)ip + hlen);
if (uh->uh_sum == 0)
return; /* no checksum */
break;
default:
return;
}
/* Extract computed checksum. */
csum = be16dec(mtod(m, char *) + pos);
/* checksum fixup for IP options */
len = hlen - sizeof(struct ip);
if (len > 0) {
opts = (uint16_t *)(ip + 1);
for (; len > 0; len -= sizeof(uint16_t), opts++) {
temp32 = csum - *opts;
temp32 = (temp32 >> 16) + (temp32 & 65535);
csum = temp32 & 65535;
}
}
m->m_pkthdr.csum_flags |= CSUM_DATA_VALID;
m->m_pkthdr.csum_data = csum;
}
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
fxp_intr_body(struct fxp_softc *sc, struct ifnet *ifp, uint8_t statack,
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
int count)
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
{
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
struct mbuf *m;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
struct fxp_rx *rxp;
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
struct fxp_rfa *rfa;
int rnr = (statack & FXP_SCB_STATACK_RNR) ? 1 : 0;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int rx_npkts;
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
uint16_t status;
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts = 0;
Abstract the locking in fxp(4) a bit more by using macros for mtx_assert() and mtx_owned(), as it is done in other places, for instance proc locking.
2004-06-02 22:59:57 +00:00
FXP_LOCK_ASSERT(sc, MA_OWNED);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
if (rnr)
Use the device sysctl tree instead of rolling our own. Some of the sysctls were global (hw.fxp_rnr and hw.fxp_noflow), all of them are now per-device. Sample output of "sysctl dev.fxp0" with this patch, with the standard %foo nodes removed : dev.fxp0.int_delay: 1000 dev.fxp0.bundle_max: 6 dev.fxp0.rnr: 0 dev.fxp0.noflow: 0
2004-06-02 22:52:18 +00:00
sc->rnr++;
Properly fix the occassional random crash issue that revision 1.142 just limited to the DEVICE_POLLING case. This removes the FXP_RFA_RNRMARK hack, and replaces it with a softc flag that is used to record when the handling of a no-resource condition was deferred due to running out of DEVICE_POLLING cycles. This was tested on -stable, but the code is essentially the same as in -current. It should only affect the case where DEVICE_POLLING is defined. The details of the mechanism behind the crashes are still uncertain but the most likely cause seems to be some kind of hardware confusion when the no-resource recovery code is accidentally invoked while the receiver is still active. This could have happened if the hardware left the 0x4000 bit of the RFA status word set. The comments in the commit log for revision 1.142 stating that the driver could clash with the hardware writing to this status word were not correct. Tested by: Guy Helmer <ghelmer@palisadesys.com>
2002-11-07 16:04:07 +00:00
#ifdef DEVICE_POLLING
/* Pick up a deferred RNR condition if `count' ran out last time. */
if (sc->flags & FXP_FLAG_DEFERRED_RNR) {
sc->flags &= ~FXP_FLAG_DEFERRED_RNR;
rnr = 1;
}
#endif
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
whitespace commit: fix indentation in fxp_intr_body. This is also done in preparation of a subsequent fix for the handling of RNR conditions in polling mode.
2002-07-29 02:48:09 +00:00
/*
* Free any finished transmit mbuf chains.
*
* Handle the CNA event likt a CXTNO event. It used to
* be that this event (control unit not ready) was not
* encountered, but it is now with the SMPng modifications.
* The exact sequence of events that occur when the interface
* is brought up are different now, and if this event
* goes unhandled, the configuration/rxfilter setup sequence
* can stall for several seconds. The result is that no
* packets go out onto the wire for about 5 to 10 seconds
* after the interface is ifconfig'ed for the first time.
*/
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
if (statack & (FXP_SCB_STATACK_CXTNO | FXP_SCB_STATACK_CNA))
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
fxp_txeof(sc);
- Allow fxp_encap() enqueue failed transmissions and set IFF_DRV_OACTIVE to note resource shortage to upper stack. - Don't count number of mbuf chains. Default 32 DMA segments for a frame is enough for most cases. If bus_dmamap_mbuf_sg fails use m_collapse(9) to collapse the mbuf chain instead of relying on expensive m_defrag(9). - Move bpf handling to fxp_start_body() which is supposed to be more appropriate place. - Always arm watchdog timer whenever a new Tx request is made. Previously fxp(4) used to arm watchdog timer only when FXP_CXINT_THRESH-th Tx request is made. Because fxp(4) does not rely on Tx interrupt to reclaim transmitted mbufs it's better to arm watchdog timer to detect potential lockups. - Add more aggresive Tx buffer reclaiming in fxp_start_body to make room for new Tx requests. Since fxp(4) does not request Tx completion interrupt for every frames it's necessary to clean TXCBs in advance to saturate link. - Make fxp(4) try to start more packets transmitting regardless of interrupt type in fxp_intr_body.
2008-11-25 04:16:16 +00:00
/*
* Try to start more packets transmitting.
*/
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
fxp_start_body(ifp);
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
/*
* Just return if nothing happened on the receive side.
*/
Properly fix the occassional random crash issue that revision 1.142 just limited to the DEVICE_POLLING case. This removes the FXP_RFA_RNRMARK hack, and replaces it with a softc flag that is used to record when the handling of a no-resource condition was deferred due to running out of DEVICE_POLLING cycles. This was tested on -stable, but the code is essentially the same as in -current. It should only affect the case where DEVICE_POLLING is defined. The details of the mechanism behind the crashes are still uncertain but the most likely cause seems to be some kind of hardware confusion when the no-resource recovery code is accidentally invoked while the receiver is still active. This could have happened if the hardware left the 0x4000 bit of the RFA status word set. The comments in the commit log for revision 1.142 stating that the driver could clash with the hardware writing to this status word were not correct. Tested by: Guy Helmer <ghelmer@palisadesys.com>
2002-11-07 16:04:07 +00:00
if (!rnr && (statack & FXP_SCB_STATACK_FR) == 0)
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
whitespace commit: fix indentation in fxp_intr_body. This is also done in preparation of a subsequent fix for the handling of RNR conditions in polling mode.
2002-07-29 02:48:09 +00:00
/*
* Process receiver interrupts. If a no-resource (RNR)
* condition exists, get whatever packets we can and
* re-start the receiver.
Properly fix the occassional random crash issue that revision 1.142 just limited to the DEVICE_POLLING case. This removes the FXP_RFA_RNRMARK hack, and replaces it with a softc flag that is used to record when the handling of a no-resource condition was deferred due to running out of DEVICE_POLLING cycles. This was tested on -stable, but the code is essentially the same as in -current. It should only affect the case where DEVICE_POLLING is defined. The details of the mechanism behind the crashes are still uncertain but the most likely cause seems to be some kind of hardware confusion when the no-resource recovery code is accidentally invoked while the receiver is still active. This could have happened if the hardware left the 0x4000 bit of the RFA status word set. The comments in the commit log for revision 1.142 stating that the driver could clash with the hardware writing to this status word were not correct. Tested by: Guy Helmer <ghelmer@palisadesys.com>
2002-11-07 16:04:07 +00:00
*
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
* When using polling, we do not process the list to completion,
* so when we get an RNR interrupt we must defer the restart
* until we hit the last buffer with the C bit set.
* If we run out of cycles and rfa_headm has the C bit set,
Properly fix the occassional random crash issue that revision 1.142 just limited to the DEVICE_POLLING case. This removes the FXP_RFA_RNRMARK hack, and replaces it with a softc flag that is used to record when the handling of a no-resource condition was deferred due to running out of DEVICE_POLLING cycles. This was tested on -stable, but the code is essentially the same as in -current. It should only affect the case where DEVICE_POLLING is defined. The details of the mechanism behind the crashes are still uncertain but the most likely cause seems to be some kind of hardware confusion when the no-resource recovery code is accidentally invoked while the receiver is still active. This could have happened if the hardware left the 0x4000 bit of the RFA status word set. The comments in the commit log for revision 1.142 stating that the driver could clash with the hardware writing to this status word were not correct. Tested by: Guy Helmer <ghelmer@palisadesys.com>
2002-11-07 16:04:07 +00:00
* record the pending RNR in the FXP_FLAG_DEFERRED_RNR flag so
* that the info will be used in the subsequent polling cycle.
whitespace commit: fix indentation in fxp_intr_body. This is also done in preparation of a subsequent fix for the handling of RNR conditions in polling mode.
2002-07-29 02:48:09 +00:00
*/
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
for (;;) {
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
rxp = sc->fxp_desc.rx_head;
m = rxp->rx_mbuf;
whitespace commit: fix indentation in fxp_intr_body. This is also done in preparation of a subsequent fix for the handling of RNR conditions in polling mode.
2002-07-29 02:48:09 +00:00
rfa = (struct fxp_rfa *)(m->m_ext.ext_buf +
RFA_ALIGNMENT_FUDGE);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_rxmtag, rxp->rx_map,
Controller updates RFA via DMA so driver needs synchronization. Add missing BUS_DMASYNC_POSTWRITE and BUS_DMASYNC_PREREAD.
2010-05-14 16:55:13 +00:00
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Device Polling code for -current. Non-SMP, i386-only, no polling in the idle loop at the moment. To use this code you must compile a kernel with options DEVICE_POLLING and at runtime enable polling with sysctl kern.polling.enable=1 The percentage of CPU reserved to userland can be set with sysctl kern.polling.user_frac=NN (default is 50) while the remainder is used by polling device drivers and netisr's. These are the only two variables that you should need to touch. There are a few more parameters in kern.polling but the default values are adequate for all purposes. See the code in kern_poll.c for more details on them. Polling in the idle loop will be implemented shortly by introducing a kernel thread which does the job. Until then, the amount of CPU dedicated to polling will never exceed (100-user_frac). The equivalent (actually, better) code for -stable is at http://info.iet.unipi.it/~luigi/polling/ and also supports polling in the idle loop. NOTE to Alpha developers: There is really nothing in this code that is i386-specific. If you move the 2 lines supporting the new option from sys/conf/{files,options}.i386 to sys/conf/{files,options} I am pretty sure that this should work on the Alpha as well, just that I do not have a suitable test box to try it. If someone feels like trying it, I would appreciate it. NOTE to other developers: sure some things could be done better, and as always I am open to constructive criticism, which a few of you have already given and I greatly appreciated. However, before proposing radical architectural changes, please take some time to possibly try out this code, or at the very least read the comments in kern_poll.c, especially re. the reason why I am using a soft netisr and cannot (I believe) replace it with a simple timeout. Quick description of files touched by this commit: sys/conf/files.i386 new file kern/kern_poll.c sys/conf/options.i386 new option sys/i386/i386/trap.c poll in trap (disabled by default) sys/kern/kern_clock.c initialization and hardclock hooks. sys/kern/kern_intr.c minor swi_net changes sys/kern/kern_poll.c the bulk of the code. sys/net/if.h new flag sys/net/if_var.h declaration for functions used in device drivers. sys/net/netisr.h NETISR_POLL sys/dev/fxp/if_fxp.c sys/dev/fxp/if_fxpvar.h sys/pci/if_dc.c sys/pci/if_dcreg.h sys/pci/if_sis.c sys/pci/if_sisreg.h device driver modifications
2001-12-14 17:56:12 +00:00
#ifdef DEVICE_POLLING /* loop at most count times if count >=0 */
Properly fix the occassional random crash issue that revision 1.142 just limited to the DEVICE_POLLING case. This removes the FXP_RFA_RNRMARK hack, and replaces it with a softc flag that is used to record when the handling of a no-resource condition was deferred due to running out of DEVICE_POLLING cycles. This was tested on -stable, but the code is essentially the same as in -current. It should only affect the case where DEVICE_POLLING is defined. The details of the mechanism behind the crashes are still uncertain but the most likely cause seems to be some kind of hardware confusion when the no-resource recovery code is accidentally invoked while the receiver is still active. This could have happened if the hardware left the 0x4000 bit of the RFA status word set. The comments in the commit log for revision 1.142 stating that the driver could clash with the hardware writing to this status word were not correct. Tested by: Guy Helmer <ghelmer@palisadesys.com>
2002-11-07 16:04:07 +00:00
if (count >= 0 && count-- == 0) {
if (rnr) {
/* Defer RNR processing until the next time. */
sc->flags |= FXP_FLAG_DEFERRED_RNR;
rnr = 0;
}
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
break;
Properly fix the occassional random crash issue that revision 1.142 just limited to the DEVICE_POLLING case. This removes the FXP_RFA_RNRMARK hack, and replaces it with a softc flag that is used to record when the handling of a no-resource condition was deferred due to running out of DEVICE_POLLING cycles. This was tested on -stable, but the code is essentially the same as in -current. It should only affect the case where DEVICE_POLLING is defined. The details of the mechanism behind the crashes are still uncertain but the most likely cause seems to be some kind of hardware confusion when the no-resource recovery code is accidentally invoked while the receiver is still active. This could have happened if the hardware left the 0x4000 bit of the RFA status word set. The comments in the commit log for revision 1.142 stating that the driver could clash with the hardware writing to this status word were not correct. Tested by: Guy Helmer <ghelmer@palisadesys.com>
2002-11-07 16:04:07 +00:00
}
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
#endif /* DEVICE_POLLING */
whitespace commit: fix indentation in fxp_intr_body. This is also done in preparation of a subsequent fix for the handling of RNR conditions in polling mode.
2002-07-29 02:48:09 +00:00
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
status = le16toh(rfa->rfa_status);
if ((status & FXP_RFA_STATUS_C) == 0)
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
break;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
/*
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
* Advance head forward.
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
*/
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
sc->fxp_desc.rx_head = rxp->rx_next;
whitespace commit: fix indentation in fxp_intr_body. This is also done in preparation of a subsequent fix for the handling of RNR conditions in polling mode.
2002-07-29 02:48:09 +00:00
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
/*
* Add a new buffer to the receive chain.
* If this fails, the old buffer is recycled
* instead.
*/
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
if (fxp_new_rfabuf(sc, rxp) == 0) {
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
int total_len;
/*
* Fetch packet length (the top 2 bits of
* actual_size are flags set by the controller
* upon completion), and drop the packet in case
* of bogus length or CRC errors.
*/
- Endian fixes. - 6 -> ETHER_ADDR_LEN.
2003-04-08 17:21:15 +00:00
total_len = le16toh(rfa->actual_size) & 0x3fff;
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
if ((sc->flags & FXP_FLAG_82559_RXCSUM) != 0 &&
(ifp->if_capenable & IFCAP_RXCSUM) != 0) {
/* Adjust for appended checksum bytes. */
total_len -= 2;
}
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
if (total_len < sizeof(struct ether_header) ||
total_len > MCLBYTES - RFA_ALIGNMENT_FUDGE -
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
sc->rfa_size || status & FXP_RFA_STATUS_CRC) {
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
m_freem(m);
continue;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
m->m_pkthdr.len = m->m_len = total_len;
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
m->m_pkthdr.rcvif = ifp;
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
/* Do IP checksum checking. */
if ((ifp->if_capenable & IFCAP_RXCSUM) != 0)
fxp_rxcsum(sc, ifp, m, status, total_len);
Add VLAN hardware tag insertion/stripping support. Tx/Rx checksum offload for VLAN frames are also supported. The VLAN hardware assistance is available only on 82550/82551 based controllers. While I'm here change the confusing name of bit1 in byte 22 of configuration block to vlan_drop_en. The bit controls whether hardware strips VLAN tagged frame or not. Special thanks to wpaul who sent valuable VLAN related information to me. Tested on: i386, sparc64
2008-12-02 02:30:12 +00:00
if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0 &&
(status & FXP_RFA_STATUS_VLAN) != 0) {
m->m_pkthdr.ether_vtag =
ntohs(rfa->rfax_vlan_id);
m->m_flags |= M_VLANTAG;
}
Drop locks before calling if_input() since it may re-enter fxp_start() in the netisr case. This would result in a lock reversal. This fixes the net.isr.enable=1 case. Better performance might be obtained by chaining all packets received, dropping the lock, and then calling if_input() on each one. Reported by: hmp
2003-06-23 23:23:49 +00:00
/*
* Drop locks before calling if_input() since it
* may re-enter fxp_start() in the netisr case.
* This would result in a lock reversal. Better
* performance might be obtained by chaining all
* packets received, dropping the lock, and then
* calling if_input() on each one.
*/
FXP_UNLOCK(sc);
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
(*ifp->if_input)(ifp, m);
Drop locks before calling if_input() since it may re-enter fxp_start() in the netisr case. This would result in a lock reversal. This fixes the net.isr.enable=1 case. Better performance might be obtained by chaining all packets received, dropping the lock, and then calling if_input() on each one. Reported by: hmp
2003-06-23 23:23:49 +00:00
FXP_LOCK(sc);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
rx_npkts++;
Always check fxp(4) is running, see if it can accept frames from upper stack in fxp_start_body(). fxp(4) drops driver lock in Rx path so check the fxp(4) is still running after reacquiring driver lock in Rx path. Also don't invoke fxp_intr_body if fxp(4) is not running. With this change there is no need to set suspend bit in device attach phase.
2009-06-21 06:46:32 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
return (rx_npkts);
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
} else {
/* Reuse RFA and loaded DMA map. */
ifp->if_iqdrops++;
fxp_discard_rfabuf(sc, rxp);
whitespace commit: fix indentation in fxp_intr_body. This is also done in preparation of a subsequent fix for the handling of RNR conditions in polling mode.
2002-07-29 02:48:09 +00:00
}
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
fxp_add_rfabuf(sc, rxp);
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
}
if (rnr) {
Properly fix the occassional random crash issue that revision 1.142 just limited to the DEVICE_POLLING case. This removes the FXP_RFA_RNRMARK hack, and replaces it with a softc flag that is used to record when the handling of a no-resource condition was deferred due to running out of DEVICE_POLLING cycles. This was tested on -stable, but the code is essentially the same as in -current. It should only affect the case where DEVICE_POLLING is defined. The details of the mechanism behind the crashes are still uncertain but the most likely cause seems to be some kind of hardware confusion when the no-resource recovery code is accidentally invoked while the receiver is still active. This could have happened if the hardware left the 0x4000 bit of the RFA status word set. The comments in the commit log for revision 1.142 stating that the driver could clash with the hardware writing to this status word were not correct. Tested by: Guy Helmer <ghelmer@palisadesys.com>
2002-11-07 16:04:07 +00:00
fxp_scb_wait(sc);
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL,
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
sc->fxp_desc.rx_head->rx_addr);
Properly fix the occassional random crash issue that revision 1.142 just limited to the DEVICE_POLLING case. This removes the FXP_RFA_RNRMARK hack, and replaces it with a softc flag that is used to record when the handling of a no-resource condition was deferred due to running out of DEVICE_POLLING cycles. This was tested on -stable, but the code is essentially the same as in -current. It should only affect the case where DEVICE_POLLING is defined. The details of the mechanism behind the crashes are still uncertain but the most likely cause seems to be some kind of hardware confusion when the no-resource recovery code is accidentally invoked while the receiver is still active. This could have happened if the hardware left the 0x4000 bit of the RFA status word set. The comments in the commit log for revision 1.142 stating that the driver could clash with the hardware writing to this status word were not correct. Tested by: Guy Helmer <ghelmer@palisadesys.com>
2002-11-07 16:04:07 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_RU_START);
whitespace commit: fix indentation in fxp_intr_body. This is also done in preparation of a subsequent fix for the handling of RNR conditions in polling mode.
2002-07-29 02:48:09 +00:00
}
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Export hardware MAC statistics through sysctl node. Previously fxp(4) already used to extract most hardware MAC statistics but it didn't show them. With this change, all MAC statistics counters are exported. Because there are a couple of new counters for 82558 and 82559, enable extended MAC statistics functionality to get these counters. Accoring to public data sheet, 82559 MAC statistics return 24 DWORD counters(3 counters are unknown at this moment) so increase MAC counter structure to meet the MAC statistics block size. The completion of MAC counter dump is now checked against FXP_STATS_DR_COMPLETE status code which is appended at the end of status block. Previously fxp(4) ignored the status of the FXP_SCB_COMMAND_CU_DUMPRESET command. fxp(4) does not wait for the completion of pending command before issuing FXP_SCB_COMMAND_CU_DUMPRESET. Instead it skips the command and try it next time. This scheme may show better performance but there is chance to loose updated counters after stopping controller. So make sure to update MAC statistics in fxp_stop(). While I'm here move sysctl node creation to fxp_sysctl_node(). Tested by: Larry Baird < lab <> gta dot com >
2010-05-09 22:16:15 +00:00
static void
fxp_update_stats(struct fxp_softc *sc)
{
struct ifnet *ifp = sc->ifp;
struct fxp_stats *sp = sc->fxp_stats;
struct fxp_hwstats *hsp;
uint32_t *status;
FXP_LOCK_ASSERT(sc, MA_OWNED);
bus_dmamap_sync(sc->fxp_stag, sc->fxp_smap,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
/* Update statistical counters. */
if (sc->revision >= FXP_REV_82559_A0)
status = &sp->completion_status;
else if (sc->revision >= FXP_REV_82558_A4)
status = (uint32_t *)&sp->tx_tco;
else
status = &sp->tx_pause;
if (*status == htole32(FXP_STATS_DR_COMPLETE)) {
hsp = &sc->fxp_hwstats;
hsp->tx_good += le32toh(sp->tx_good);
hsp->tx_maxcols += le32toh(sp->tx_maxcols);
hsp->tx_latecols += le32toh(sp->tx_latecols);
hsp->tx_underruns += le32toh(sp->tx_underruns);
hsp->tx_lostcrs += le32toh(sp->tx_lostcrs);
hsp->tx_deffered += le32toh(sp->tx_deffered);
hsp->tx_single_collisions += le32toh(sp->tx_single_collisions);
hsp->tx_multiple_collisions +=
le32toh(sp->tx_multiple_collisions);
hsp->tx_total_collisions += le32toh(sp->tx_total_collisions);
hsp->rx_good += le32toh(sp->rx_good);
hsp->rx_crc_errors += le32toh(sp->rx_crc_errors);
hsp->rx_alignment_errors += le32toh(sp->rx_alignment_errors);
hsp->rx_rnr_errors += le32toh(sp->rx_rnr_errors);
hsp->rx_overrun_errors += le32toh(sp->rx_overrun_errors);
hsp->rx_cdt_errors += le32toh(sp->rx_cdt_errors);
hsp->rx_shortframes += le32toh(sp->rx_shortframes);
hsp->tx_pause += le32toh(sp->tx_pause);
hsp->rx_pause += le32toh(sp->rx_pause);
hsp->rx_controls += le32toh(sp->rx_controls);
hsp->tx_tco += le16toh(sp->tx_tco);
hsp->rx_tco += le16toh(sp->rx_tco);
ifp->if_opackets += le32toh(sp->tx_good);
ifp->if_collisions += le32toh(sp->tx_total_collisions);
if (sp->rx_good) {
ifp->if_ipackets += le32toh(sp->rx_good);
sc->rx_idle_secs = 0;
} else if (sc->flags & FXP_FLAG_RXBUG) {
/*
* Receiver's been idle for another second.
*/
sc->rx_idle_secs++;
}
ifp->if_ierrors +=
le32toh(sp->rx_crc_errors) +
le32toh(sp->rx_alignment_errors) +
le32toh(sp->rx_rnr_errors) +
le32toh(sp->rx_overrun_errors);
/*
* If any transmit underruns occured, bump up the transmit
* threshold by another 512 bytes (64 * 8).
*/
if (sp->tx_underruns) {
ifp->if_oerrors += le32toh(sp->tx_underruns);
if (tx_threshold < 192)
tx_threshold += 64;
}
*status = 0;
bus_dmamap_sync(sc->fxp_stag, sc->fxp_smap,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
}
}
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
/*
* Update packet in/out/collision statistics. The i82557 doesn't
* allow you to access these counters without doing a fairly
* expensive DMA to get _all_ of the statistics it maintains, so
* we do this operation here only once per second. The statistics
* counters in the kernel are updated from the previous dump-stats
* DMA and then a new dump-stats DMA is started. The on-chip
* counters are zeroed when the DMA completes. If we can't start
* the DMA immediately, we don't wait - we just prepare to read
* them again next time.
*/
Staticize.
1998-02-09 06:11:36 +00:00
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_tick(void *xsc)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
struct fxp_softc *sc = xsc;
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
struct ifnet *ifp = sc->ifp;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
FXP_LOCK_ASSERT(sc, MA_OWNED);
Export hardware MAC statistics through sysctl node. Previously fxp(4) already used to extract most hardware MAC statistics but it didn't show them. With this change, all MAC statistics counters are exported. Because there are a couple of new counters for 82558 and 82559, enable extended MAC statistics functionality to get these counters. Accoring to public data sheet, 82559 MAC statistics return 24 DWORD counters(3 counters are unknown at this moment) so increase MAC counter structure to meet the MAC statistics block size. The completion of MAC counter dump is now checked against FXP_STATS_DR_COMPLETE status code which is appended at the end of status block. Previously fxp(4) ignored the status of the FXP_SCB_COMMAND_CU_DUMPRESET command. fxp(4) does not wait for the completion of pending command before issuing FXP_SCB_COMMAND_CU_DUMPRESET. Instead it skips the command and try it next time. This scheme may show better performance but there is chance to loose updated counters after stopping controller. So make sure to update MAC statistics in fxp_stop(). While I'm here move sysctl node creation to fxp_sysctl_node(). Tested by: Larry Baird < lab <> gta dot com >
2010-05-09 22:16:15 +00:00
/* Update statistical counters. */
fxp_update_stats(sc);
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
Defer rundown (m_freem) of completed transmit buffers for no longer than 1 second.
1998-10-22 02:00:49 +00:00
/*
* Release any xmit buffers that have completed DMA. This isn't
* strictly necessary to do here, but it's advantagous for mbufs
* with external storage to be released in a timely manner rather
* than being defered for a potentially long time. This limits
* the delay to a maximum of one second.
Whitespace fix.
2008-11-25 00:48:05 +00:00
*/
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
fxp_txeof(sc);
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
/*
* If we haven't received any packets in FXP_MAC_RX_IDLE seconds,
* then assume the receiver has locked up and attempt to clear
* the condition by reprogramming the multicast filter. This is
* a work-around for a bug in the 82557 where the receiver locks
* up if it gets certain types of garbage in the syncronization
* bits prior to the packet header. This bug is supposed to only
* occur in 10Mbps mode, but has been seen to occur in 100Mbps
* mode as well (perhaps due to a 10/100 speed transition).
*/
if (sc->rx_idle_secs > FXP_MAX_RX_IDLE) {
sc->rx_idle_secs = 0;
Always check fxp(4) is running, see if it can accept frames from upper stack in fxp_start_body(). fxp(4) drops driver lock in Rx path so check the fxp(4) is still running after reacquiring driver lock in Rx path. Also don't invoke fxp_intr_body if fxp(4) is not running. With this change there is no need to set suspend bit in device attach phase.
2009-06-21 06:46:32 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
fxp_init_body(sc);
return;
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
}
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
* If there is no pending command, start another stats
* dump. Otherwise punt for now.
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
*/
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
if (CSR_READ_1(sc, FXP_CSR_SCB_COMMAND) == 0) {
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
/*
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
* Start another stats dump.
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
*/
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_DUMPRESET);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
if (sc->miibus != NULL)
mii_tick(device_get_softc(sc->miibus));
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on fxp_tick() callout.
2006-11-30 14:58:01 +00:00
/*
Grammar nit.
2006-12-01 13:07:52 +00:00
* Check that chip hasn't hung.
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on fxp_tick() callout.
2006-11-30 14:58:01 +00:00
*/
fxp_watchdog(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* Schedule another timeout one second from now.
*/
change timeout to be MPSAFE Sponsored by: FreeBSD Foundation
2003-09-05 22:37:31 +00:00
callout_reset(&sc->stat_ch, hz, fxp_tick, sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
/*
* Stop the interface. Cancels the statistics updater and resets
* the interface.
*/
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_stop(struct fxp_softc *sc)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
struct fxp_tx *txp;
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
int i;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +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 | IFF_DRV_OACTIVE);
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on fxp_tick() callout.
2006-11-30 14:58:01 +00:00
sc->watchdog_timer = 0;
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* Cancel stats updater.
*/
change timeout to be MPSAFE Sponsored by: FreeBSD Foundation
2003-09-05 22:37:31 +00:00
callout_stop(&sc->stat_ch);
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
/*
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
* Preserve PCI configuration, configure, IA/multicast
* setup and put RU and CU into idle state.
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
*/
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SELECTIVE_RESET);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
DELAY(50);
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
/* Disable interrupts. */
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, FXP_SCB_INTR_DISABLE);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Export hardware MAC statistics through sysctl node. Previously fxp(4) already used to extract most hardware MAC statistics but it didn't show them. With this change, all MAC statistics counters are exported. Because there are a couple of new counters for 82558 and 82559, enable extended MAC statistics functionality to get these counters. Accoring to public data sheet, 82559 MAC statistics return 24 DWORD counters(3 counters are unknown at this moment) so increase MAC counter structure to meet the MAC statistics block size. The completion of MAC counter dump is now checked against FXP_STATS_DR_COMPLETE status code which is appended at the end of status block. Previously fxp(4) ignored the status of the FXP_SCB_COMMAND_CU_DUMPRESET command. fxp(4) does not wait for the completion of pending command before issuing FXP_SCB_COMMAND_CU_DUMPRESET. Instead it skips the command and try it next time. This scheme may show better performance but there is chance to loose updated counters after stopping controller. So make sure to update MAC statistics in fxp_stop(). While I'm here move sysctl node creation to fxp_sysctl_node(). Tested by: Larry Baird < lab <> gta dot com >
2010-05-09 22:16:15 +00:00
fxp_update_stats(sc);
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
/*
* Release any xmit buffers.
*/
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
txp = sc->fxp_desc.tx_list;
Fixed mbuf leak in fxp_stop().
1998-10-10 19:26:40 +00:00
if (txp != NULL) {
for (i = 0; i < FXP_NTXCB; i++) {
Fix 5 bugs: 1) always call fxp_stop in fxp_detach. Since we don't read from the card, there's no need to carefully look at things with bus_child_present. 2) Call FXP_UNLOCK() before calling bus_teardown_intr to avoid a possible deadlock reported by jhb. 3) add gone to the softc. Set it to true in detach. 4) Return immediately if gone is true in fxp_ioctl 5) Return immediately if gone is true in fxp_intr
2003-04-29 05:45:09 +00:00
if (txp[i].tx_mbuf != NULL) {
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_txmtag, txp[i].tx_map,
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
BUS_DMASYNC_POSTWRITE);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_unload(sc->fxp_txmtag,
txp[i].tx_map);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
m_freem(txp[i].tx_mbuf);
txp[i].tx_mbuf = NULL;
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
/* clear this to reset csum offload bits */
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
txp[i].tx_cb->tbd[0].tb_addr = 0;
Fixed mbuf leak in fxp_stop().
1998-10-10 19:26:40 +00:00
}
}
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
}
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
sc->tx_queued = 0;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
/*
* Watchdog/transmission transmit timeout handler. Called when a
* transmission is started on the interface, but no interrupt is
* received before the timeout. This usually indicates that the
* card has wedged for some reason.
*/
static void
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on fxp_tick() callout.
2006-11-30 14:58:01 +00:00
fxp_watchdog(struct fxp_softc *sc)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on fxp_tick() callout.
2006-11-30 14:58:01 +00:00
FXP_LOCK_ASSERT(sc, MA_OWNED);
if (sc->watchdog_timer == 0 || --sc->watchdog_timer)
return;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
device_printf(sc->dev, "device timeout\n");
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on fxp_tick() callout.
2006-11-30 14:58:01 +00:00
sc->ifp->if_oerrors++;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
fxp_init_body(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
/*
* Acquire locks and then call the real initialization function. This
* is necessary because ether_ioctl() calls if_init() and this would
* result in mutex recursion if the mutex was held.
*/
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_init(void *xsc)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
1) Implement SIOCSIFMTU in ether_ioctl(), and change ether_ioctl's return type to be int so that errors can be returned. 2) Use the new SIOCSIFMTU ether_ioctl support in the few drivers that are using ether_ioctl(). 3) In if_fxp.c: treat if_bpf as a token, not as a pointer. Don't bother testing for FXP_NTXSEG being reached in fxp_start()...just check for non-NULL 'm'. Change fxp_ioctl() to use ether_ioctl().
1996-12-10 07:29:50 +00:00
struct fxp_softc *sc = xsc;
Make fxp(4) INTR_MPSAFE (but do not enable MPSAFE just yet): - Add fxp_start_body() and change fxp_start() to just acquire locks and then call fxp_start_body(). Places that would call fxp_start() with locks held (mutex recursion) now call fxp_start_body() directly. Remove MTX_RECURSE flag from sc_mtx. [gallatin] - Change fxp_attach() to work without the softc lock, saving interrupt hooking until the head of fxp_attach(). - Call ether_ifattach() before overriding ifp parameters. This reverts part of 1.155. - Remove multiple error paths in fxp_attach(). - Teardown interrupt in fxp_detach() before unlocking the softc. - Make sure mutex is not held in fxp_release() - Delete the miibus instance and/or self in fxp_release(), not in fxp_detach(). This can happen if attach fails partway through. - Move ifmedia_removeall to fxp_release() since attach may fail after media have been allocated. - Add locking to fxp_suspend, fxp_resume, fxp_start, fxp_intr, fxp_poll, fxp_tick, fxp_ioctl, fxp_watchdog. - Pass in ifp to fxp_intr_body since its callers sometimes already use it. - Add compatibility define for INTR_MPSAFE for 4.x. [gallatin] - You don't need to bzero softc. Ideas from: gallatin, mux Tested by: >400M packets of dd/ssh, NFS, ping on i386 UP
2003-04-25 09:01:54 +00:00
FXP_LOCK(sc);
fxp_init_body(sc);
FXP_UNLOCK(sc);
}
/*
* Perform device initialization. This routine must be called with the
* softc lock held.
*/
static void
fxp_init_body(struct fxp_softc *sc)
{
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
struct ifnet *ifp = sc->ifp;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
struct fxp_cb_config *cbp;
struct fxp_cb_ias *cb_ias;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
struct fxp_cb_tx *tcbp;
struct fxp_tx *txp;
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
int i, prm;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Abstract the locking in fxp(4) a bit more by using macros for mtx_assert() and mtx_owned(), as it is done in other places, for instance proc locking.
2004-06-02 22:59:57 +00:00
FXP_LOCK_ASSERT(sc, MA_OWNED);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
* Cancel any pending I/O
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
*/
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
fxp_stop(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
/*
* Issue software reset, which also unloads the microcode.
*/
sc->flags &= ~FXP_FLAG_UCODE;
CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SOFTWARE_RESET);
DELAY(50);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
prm = (ifp->if_flags & IFF_PROMISC) ? 1 : 0;
/*
* Initialize base of CBL and RFA memory. Loading with zero
* sets it up for regular linear addressing.
*/
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, 0);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_BASE);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
fxp_scb_wait(sc);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_RU_BASE);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* Initialize base of dump-stats buffer.
*/
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
fxp_scb_wait(sc);
Export hardware MAC statistics through sysctl node. Previously fxp(4) already used to extract most hardware MAC statistics but it didn't show them. With this change, all MAC statistics counters are exported. Because there are a couple of new counters for 82558 and 82559, enable extended MAC statistics functionality to get these counters. Accoring to public data sheet, 82559 MAC statistics return 24 DWORD counters(3 counters are unknown at this moment) so increase MAC counter structure to meet the MAC statistics block size. The completion of MAC counter dump is now checked against FXP_STATS_DR_COMPLETE status code which is appended at the end of status block. Previously fxp(4) ignored the status of the FXP_SCB_COMMAND_CU_DUMPRESET command. fxp(4) does not wait for the completion of pending command before issuing FXP_SCB_COMMAND_CU_DUMPRESET. Instead it skips the command and try it next time. This scheme may show better performance but there is chance to loose updated counters after stopping controller. So make sure to update MAC statistics in fxp_stop(). While I'm here move sysctl node creation to fxp_sysctl_node(). Tested by: Larry Baird < lab <> gta dot com >
2010-05-09 22:16:15 +00:00
bzero(sc->fxp_stats, sizeof(struct fxp_stats));
bus_dmamap_sync(sc->fxp_stag, sc->fxp_smap,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->stats_addr);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_DUMP_ADR);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
/*
* Attempt to load microcode if requested.
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
* For ICH based controllers do not load microcode.
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
*/
For ICH based fxp(4) controllers treat them as 82559 compatibles. To detect which controller is ICH based one, add a new member variable ich to struct fxp_ident and move the struct to if_fxpvar.h. Since I've faked controller revision, don't allow microcode loading for ICH based controllers. With this change all ICH based controllers will have WOL and Rx checksum offload capability. PR: kern/135451 Tested by: Alexey Shuvaev ( shuvaev <> physik dot uni-wuerzburg dot de ), pluknet ( pluknet <> gmail dot com ), Gary Jennejohn ( gary.jennejohn <> freenet dot de )
2009-06-21 07:34:12 +00:00
if (sc->ident->ich == 0) {
if (ifp->if_flags & IFF_LINK0 &&
(sc->flags & FXP_FLAG_UCODE) == 0)
fxp_load_ucode(sc);
}
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
/*
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
* Set IFF_ALLMULTI status. It's needed in configure action
* command.
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
*/
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
fxp_mc_addrs(sc);
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* We temporarily use memory that contains the TxCB list to
* construct the config CB. The TxCB list memory is rebuilt
* later.
*/
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
cbp = (struct fxp_cb_config *)sc->fxp_desc.cbl_list;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* This bcopy is kind of disgusting, but there are a bunch of must be
* zero and must be one bits in this structure and this is the easiest
* way to initialize them all to proper values.
*/
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
bcopy(fxp_cb_config_template, cbp, sizeof(fxp_cb_config_template));
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cbp->cb_status = 0;
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
cbp->cb_command = htole16(FXP_CB_COMMAND_CONFIG |
FXP_CB_COMMAND_EL);
cbp->link_addr = 0xffffffff; /* (no) next command */
Remove an erroneous htole16(). The byte_count field in the struct fxp_cb_config is a 6-bit field, not a 16-bit integer as in struct fxp_cb_tx.
2003-04-07 14:53:30 +00:00
cbp->byte_count = sc->flags & FXP_FLAG_EXT_RFA ? 32 : 22;
Changed several configuration options: Disabled the DMA byte counters - I had it this way originally and this is the recommended setting. Set crscdt to CRS only (0) since this is what it should be for an MII PHY. Also fixed some comments.
1997-02-04 11:44:15 +00:00
cbp->rx_fifo_limit = 8; /* rx fifo threshold (32 bytes) */
cbp->tx_fifo_limit = 0; /* tx fifo threshold (0 bytes) */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cbp->adaptive_ifs = 0; /* (no) adaptive interframe spacing */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->mwi_enable = sc->flags & FXP_FLAG_MWI_ENABLE ? 1 : 0;
cbp->type_enable = 0; /* actually reserved */
cbp->read_align_en = sc->flags & FXP_FLAG_READ_ALIGN ? 1 : 0;
cbp->end_wr_on_cl = sc->flags & FXP_FLAG_WRITE_ALIGN ? 1 : 0;
Changed several configuration options: Disabled the DMA byte counters - I had it this way originally and this is the recommended setting. Set crscdt to CRS only (0) since this is what it should be for an MII PHY. Also fixed some comments.
1997-02-04 11:44:15 +00:00
cbp->rx_dma_bytecount = 0; /* (no) rx DMA max */
cbp->tx_dma_bytecount = 0; /* (no) tx DMA max */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->dma_mbce = 0; /* (disable) dma max counters */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cbp->late_scb = 0; /* (don't) defer SCB update */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->direct_dma_dis = 1; /* disable direct rcv dma mode */
cbp->tno_int_or_tco_en =0; /* (disable) tx not okay interrupt */
Implemented a more sophisticated mechanism for handling transmitter interrupts which now defers them until the transmit queue if filled up with completed buffers. This has two advantages: first, it reduces the number of transmitter interrupts to just 1/120th of the rate that they occured previously, and two, running down many buffers at once has much improved cache effects.
1998-08-04 08:53:12 +00:00
cbp->ci_int = 1; /* interrupt on CU idle */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->ext_txcb_dis = sc->flags & FXP_FLAG_EXT_TXCB ? 0 : 1;
cbp->ext_stats_dis = 1; /* disable extended counters */
cbp->keep_overrun_rx = 0; /* don't pass overrun frames to host */
Teach fxp(4) to control VLAN_MTU in the hardware. Now reception of extended frames can be toggled through ioctl(SIOCSIFCAP). The card will also receive extended frames when in promiscuous mode.
2004-05-25 14:49:46 +00:00
cbp->save_bf = sc->flags & FXP_FLAG_SAVE_BAD ? 1 : prm;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cbp->disc_short_rx = !prm; /* discard short packets */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->underrun_retry = 1; /* retry mode (once) on DMA underrun */
cbp->two_frames = 0; /* do not limit FIFO to 2 frames */
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
cbp->dyn_tbd = sc->flags & FXP_FLAG_EXT_RFA ? 1 : 0;
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
cbp->ext_rfa = sc->flags & FXP_FLAG_EXT_RFA ? 1 : 0;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->mediatype = sc->flags & FXP_FLAG_SERIAL_MEDIA ? 0 : 1;
cbp->csma_dis = 0; /* (don't) disable link */
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
cbp->tcp_udp_cksum = ((sc->flags & FXP_FLAG_82559_RXCSUM) != 0 &&
(ifp->if_capenable & IFCAP_RXCSUM) != 0) ? 1 : 0;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->vlan_tco = 0; /* (don't) enable vlan wakeup */
cbp->link_wake_en = 0; /* (don't) assert PME# on link change */
cbp->arp_wake_en = 0; /* (don't) assert PME# on arp */
cbp->mc_wake_en = 0; /* (don't) enable PME# on mcmatch */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cbp->nsai = 1; /* (don't) disable source addr insert */
cbp->preamble_length = 2; /* (7 byte) preamble */
cbp->loopback = 0; /* (don't) loopback */
cbp->linear_priority = 0; /* (normal CSMA/CD operation) */
cbp->linear_pri_mode = 0; /* (wait after xmit only) */
cbp->interfrm_spacing = 6; /* (96 bits of) interframe spacing */
cbp->promiscuous = prm; /* promiscuous mode */
cbp->bcast_disable = 0; /* (don't) disable broadcasts */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->wait_after_win = 0; /* (don't) enable modified backoff alg*/
cbp->ignore_ul = 0; /* consider U/L bit in IA matching */
cbp->crc16_en = 0; /* (don't) enable crc-16 algorithm */
cbp->crscdt = sc->flags & FXP_FLAG_SERIAL_MEDIA ? 1 : 0;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cbp->stripping = !prm; /* truncate rx packet to byte count */
cbp->padding = 1; /* (do) pad short tx packets */
cbp->rcv_crc_xfer = 0; /* (don't) xfer CRC to host */
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
cbp->long_rx_en = sc->flags & FXP_FLAG_LONG_PKT_EN ? 1 : 0;
cbp->ia_wake_en = 0; /* (don't) wake up on address match */
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
cbp->magic_pkt_dis = sc->flags & FXP_FLAG_WOL ? 0 : 1;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cbp->force_fdx = 0; /* (don't) force full duplex */
Reworked the stop/reset code to properly run-down allocated buffers and made the code detect and handle wedged cards.
1996-01-03 05:22:32 +00:00
cbp->fdx_pin_en = 1; /* (enable) FDX# pin */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cbp->multi_ia = 0; /* (don't) accept multiple IAs */
It seems controller has two types of promiscuous control, one for unicast and the other for multicast. To receive multicast frames that host didn't join in promiscuous mode, driver have to set promiscuous mode for multicast frames as well. The Open Source Software Developer Manual for i8255x was not clear how to handle promiscuous mode. PR: kern/145905 MFC after: 5 days
2010-05-07 16:43:00 +00:00
cbp->mc_all = ifp->if_flags & IFF_ALLMULTI ? 1 : prm;
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
cbp->gamla_rx = sc->flags & FXP_FLAG_EXT_RFA ? 1 : 0;
Add VLAN hardware tag insertion/stripping support. Tx/Rx checksum offload for VLAN frames are also supported. The VLAN hardware assistance is available only on 82550/82551 based controllers. While I'm here change the confusing name of bit1 in byte 22 of configuration block to vlan_drop_en. The bit controls whether hardware strips VLAN tagged frame or not. Special thanks to wpaul who sent valuable VLAN related information to me. Tested on: i386, sparc64
2008-12-02 02:30:12 +00:00
cbp->vlan_strip_en = ((sc->flags & FXP_FLAG_EXT_RFA) != 0 &&
(ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) ? 1 : 0;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Use the device sysctl tree instead of rolling our own. Some of the sysctls were global (hw.fxp_rnr and hw.fxp_noflow), all of them are now per-device. Sample output of "sysctl dev.fxp0" with this patch, with the standard %foo nodes removed : dev.fxp0.int_delay: 1000 dev.fxp0.bundle_max: 6 dev.fxp0.rnr: 0 dev.fxp0.noflow: 0
2004-06-02 22:52:18 +00:00
if (sc->tunable_noflow || sc->revision == FXP_REV_82557) {
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
/*
* The 82557 has no hardware flow control, the values
* below are the defaults for the chip.
*/
cbp->fc_delay_lsb = 0;
cbp->fc_delay_msb = 0x40;
cbp->pri_fc_thresh = 3;
cbp->tx_fc_dis = 0;
cbp->rx_fc_restop = 0;
cbp->rx_fc_restart = 0;
cbp->fc_filter = 0;
cbp->pri_fc_loc = 1;
} else {
cbp->fc_delay_lsb = 0x1f;
cbp->fc_delay_msb = 0x01;
cbp->pri_fc_thresh = 3;
cbp->tx_fc_dis = 0; /* enable transmit FC */
cbp->rx_fc_restop = 1; /* enable FC restop frames */
cbp->rx_fc_restart = 1; /* enable FC restart frames */
cbp->fc_filter = !prm; /* drop FC frames to host */
cbp->pri_fc_loc = 1; /* FC pri location (byte31) */
}
Export hardware MAC statistics through sysctl node. Previously fxp(4) already used to extract most hardware MAC statistics but it didn't show them. With this change, all MAC statistics counters are exported. Because there are a couple of new counters for 82558 and 82559, enable extended MAC statistics functionality to get these counters. Accoring to public data sheet, 82559 MAC statistics return 24 DWORD counters(3 counters are unknown at this moment) so increase MAC counter structure to meet the MAC statistics block size. The completion of MAC counter dump is now checked against FXP_STATS_DR_COMPLETE status code which is appended at the end of status block. Previously fxp(4) ignored the status of the FXP_SCB_COMMAND_CU_DUMPRESET command. fxp(4) does not wait for the completion of pending command before issuing FXP_SCB_COMMAND_CU_DUMPRESET. Instead it skips the command and try it next time. This scheme may show better performance but there is chance to loose updated counters after stopping controller. So make sure to update MAC statistics in fxp_stop(). While I'm here move sysctl node creation to fxp_sysctl_node(). Tested by: Larry Baird < lab <> gta dot com >
2010-05-09 22:16:15 +00:00
/* Enable 82558 and 82559 extended statistics functionality. */
if (sc->revision >= FXP_REV_82558_A4) {
if (sc->revision >= FXP_REV_82559_A0) {
/*
* Extend configuration table size to 32
* to include TCO configuration.
*/
cbp->byte_count = 32;
cbp->ext_stats_dis = 1;
/* Enable TCO stats. */
cbp->tno_int_or_tco_en = 1;
cbp->gamla_rx = 1;
} else
cbp->ext_stats_dis = 0;
}
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* Start the config command/DMA.
*/
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
fxp_scb_wait(sc);
Controller will dma SCB command status for a given command and driver should read updated status back after issuing a SCB command. To send a command to controller and read updated status back, driver should synchronize both memory read and write operations with device. Fix bus_dmamap_sync operation specifier used in fxp_dma_wait() by adding both memory read and memory write operations.
2009-06-10 01:15:30 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->fxp_desc.cbl_addr);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/* ...and wait for it to complete. */
- Pass a busdma tag and a busdma map to fxp_dma_wait() so that it can do the necessary bus_dmamap_sync() calls. - While I'm here, reorder arguments to match other functions.
2003-04-08 18:56:45 +00:00
fxp_dma_wait(sc, &cbp->cb_status, sc->cbl_tag, sc->cbl_map);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* Now initialize the station address. Temporarily use the TxCB
* memory area like we did above for the config CB.
*/
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
cb_ias = (struct fxp_cb_ias *)sc->fxp_desc.cbl_list;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
cb_ias->cb_status = 0;
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
cb_ias->cb_command = htole16(FXP_CB_COMMAND_IAS | FXP_CB_COMMAND_EL);
cb_ias->link_addr = 0xffffffff;
- Store pointer to the link-level address right in "struct ifnet" rather than in ifindex_table[]; all (except one) accesses are through ifp anyway. IF_LLADDR() works faster, and all (except one) ifaddr_byindex() users were converted to use ifp->if_addr. - Stop storing a (pointer to) Ethernet address in "struct arpcom", and drop the IFP2ENADDR() macro; all users have been converted to use IF_LLADDR() instead.
2005-11-11 16:04:59 +00:00
bcopy(IF_LLADDR(sc->ifp), cb_ias->macaddr, ETHER_ADDR_LEN);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* Start the IAS (Individual Address Setup) command/DMA.
*/
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
fxp_scb_wait(sc);
Controller will dma SCB command status for a given command and driver should read updated status back after issuing a SCB command. To send a command to controller and read updated status back, driver should synchronize both memory read and write operations with device. Fix bus_dmamap_sync operation specifier used in fxp_dma_wait() by adding both memory read and memory write operations.
2009-06-10 01:15:30 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->fxp_desc.cbl_addr);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/* ...and wait for it to complete. */
- Pass a busdma tag and a busdma map to fxp_dma_wait() so that it can do the necessary bus_dmamap_sync() calls. - While I'm here, reorder arguments to match other functions.
2003-04-08 18:56:45 +00:00
fxp_dma_wait(sc, &cb_ias->cb_status, sc->cbl_tag, sc->cbl_map);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
/*
* Initialize the multicast address list.
*/
fxp_mc_setup(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* Initialize transmit control block (TxCB) list.
*/
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
txp = sc->fxp_desc.tx_list;
tcbp = sc->fxp_desc.cbl_list;
bzero(tcbp, FXP_TXCB_SZ);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
for (i = 0; i < FXP_NTXCB; i++) {
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
txp[i].tx_mbuf = NULL;
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
tcbp[i].cb_status = htole16(FXP_CB_STATUS_C | FXP_CB_STATUS_OK);
tcbp[i].cb_command = htole16(FXP_CB_COMMAND_NOP);
tcbp[i].link_addr = htole32(sc->fxp_desc.cbl_addr +
(((i + 1) & FXP_TXCB_MASK) * sizeof(struct fxp_cb_tx)));
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
if (sc->flags & FXP_FLAG_EXT_TXCB)
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
tcbp[i].tbd_array_addr =
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
htole32(FXP_TXCB_DMA_ADDR(sc, &tcbp[i].tbd[2]));
Add some performance features to the fxp driver. If the chip is not a 82557 (e.g.: a newer chip) then: + enable MWI, if the PCI configuration indicates the system supports it + enable usage of extended TxCB, for better performance + enable hardware flow control. FC frames will be passed up to the host only if promiscuous mode is enabled.
2001-03-14 19:50:35 +00:00
else
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
tcbp[i].tbd_array_addr =
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
htole32(FXP_TXCB_DMA_ADDR(sc, &tcbp[i].tbd[0]));
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
txp[i].tx_next = &txp[(i + 1) & FXP_TXCB_MASK];
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
/*
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
* Set the suspend flag on the first TxCB and start the control
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
* unit. It will execute the NOP and then suspend.
*/
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
tcbp->cb_command = htole16(FXP_CB_COMMAND_NOP | FXP_CB_COMMAND_S);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
sc->fxp_desc.tx_first = sc->fxp_desc.tx_last = txp;
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
sc->tx_queued = 1;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
fxp_scb_wait(sc);
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->fxp_desc.cbl_addr);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* Initialize receiver buffer area - RFA.
*/
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
fxp_scb_wait(sc);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->fxp_desc.rx_head->rx_addr);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_RU_START);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
/*
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
* Set current media.
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
*/
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
if (sc->miibus != NULL)
mii_mediachg(device_get_softc(sc->miibus));
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +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;
Add few cosmetic style fixes, and some debug information for SCB timeouts. Add VLAN support, obtained from Pedro J. Lobo (through Mike Tancsa).
2001-05-13 00:03:39 +00:00
/*
* Enable interrupts.
*/
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
#ifdef DEVICE_POLLING
/*
* ... but only do that if we are not polling. And because (presumably)
* the default is interrupts on, we need to disable them explicitly!
*/
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
if (ifp->if_capenable & IFCAP_POLLING )
Fix handling of Receiver Not Ready errors when doing polling. Also take this chance to cleanup the code in fxp_intr_body. Add a missing block of code to disable interrupts when reinitializing the interface while doing polling (the RELENG_4 version was correct). MFC after: 3 days
2002-08-04 22:33:28 +00:00
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, FXP_SCB_INTR_DISABLE);
else
#endif /* DEVICE_POLLING */
Add few cosmetic style fixes, and some debug information for SCB timeouts. Add VLAN support, obtained from Pedro J. Lobo (through Mike Tancsa).
2001-05-13 00:03:39 +00:00
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, 0);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
/*
* Start stats updater.
*/
change timeout to be MPSAFE Sponsored by: FreeBSD Foundation
2003-09-05 22:37:31 +00:00
callout_reset(&sc->stat_ch, hz, fxp_tick, sc);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
}
static int
fxp_serial_ifmedia_upd(struct ifnet *ifp)
{
return (0);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
}
Staticize.
1998-02-09 06:11:36 +00:00
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_serial_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
{
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
ifmr->ifm_active = IFM_ETHER|IFM_MANUAL;
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
}
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
/*
* Change media according to request.
*/
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static int
fxp_ifmedia_upd(struct ifnet *ifp)
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
{
struct fxp_softc *sc = ifp->if_softc;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
struct mii_data *mii;
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
mii = device_get_softc(sc->miibus);
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
FXP_LOCK(sc);
Whitespace fix.
2008-11-25 00:48:05 +00:00
if (mii->mii_instance) {
struct mii_softc *miisc;
LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
mii_phy_reset(miisc);
}
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
mii_mediachg(mii);
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
FXP_UNLOCK(sc);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
return (0);
}
/*
* Notify the world which media we're using.
*/
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
static void
fxp_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
{
struct fxp_softc *sc = ifp->if_softc;
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
struct mii_data *mii;
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
mii = device_get_softc(sc->miibus);
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
FXP_LOCK(sc);
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
mii_pollstat(mii);
ifmr->ifm_active = mii->mii_media_active;
ifmr->ifm_status = mii->mii_media_status;
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
Check for 10BaseT media correctly. Before we were confusing ifm_status and ifm_active. IFM_10_T gets set in the ifm_active field, not in the ifm_status field, as far as I can tell. Note: this was to enable a workaround that's rarely enabled. I don't know how to corrupt my eeprom to test it, and would rather not know...
2006-01-04 23:00:01 +00:00
if (IFM_SUBTYPE(ifmr->ifm_active) == IFM_10_T &&
sc->flags & FXP_FLAG_CU_RESUME_BUG)
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
sc->cu_resume_bug = 1;
else
sc->cu_resume_bug = 0;
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
FXP_UNLOCK(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
/*
* Add a buffer to the end of the RFA buffer list.
* Return 0 if successful, 1 for failure. A failure results in
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
* reusing the RFA buffer.
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
* The RFA struct is stuck at the beginning of mbuf cluster and the
* data pointer is fixed up to point just past it.
*/
static int
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
fxp_new_rfabuf(struct fxp_softc *sc, struct fxp_rx *rxp)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
struct mbuf *m;
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
struct fxp_rfa *rfa;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
bus_dmamap_t tmp_map;
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
int error;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
if (m == NULL)
return (ENOBUFS);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
/*
* Move the data pointer up so that the incoming data packet
* will be 32-bit aligned.
*/
m->m_data += RFA_ALIGNMENT_FUDGE;
Fixed a bug in the receive buffer allocation code that resulted in a panic if an mbuf cluster couldn't be allocated. This was caused by a failure to re-initialize m_data when the old mbuf/mbcluster was recycled.
1996-09-20 11:05:39 +00:00
/*
* Get a pointer to the base of the mbuf cluster and move
* data start past it.
*/
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
rfa = mtod(m, struct fxp_rfa *);
As previously threatened, add TCP/IP checksum offload support to the fxp driver. This is enabled only for the 82550/82551 chips (PCI revision code 12 or 13). RX and TX checksum offload are both supported. Transmit offload is limited to TCP and UDP only right now: there seems to be a problem with IP header checksumming on transmit in some cases. This chip has hardware VLAN support as well. I hope to enable support for this eventually.
2003-02-26 22:12:04 +00:00
m->m_data += sc->rfa_size;
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
rfa->size = htole16(MCLBYTES - sc->rfa_size - RFA_ALIGNMENT_FUDGE);
Fixed a bug in the receive buffer allocation code that resulted in a panic if an mbuf cluster couldn't be allocated. This was caused by a failure to re-initialize m_data when the old mbuf/mbcluster was recycled.
1996-09-20 11:05:39 +00:00
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
rfa->rfa_status = 0;
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
rfa->rfa_control = htole16(FXP_RFA_CONTROL_EL);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
rfa->actual_size = 0;
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
m->m_len = m->m_pkthdr.len = MCLBYTES - RFA_ALIGNMENT_FUDGE -
sc->rfa_size;
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
De-inline some functions. It doesn't gain us anything and bloats code size by 3616 bytes. Furthemore, it was previously ignored by GCC. While I'm at it, fix some bogus comments.
2003-07-22 15:50:54 +00:00
/*
* Initialize the rest of the RFA. Note that since the RFA
* is misaligned, we cannot store values directly. We're thus
* using the le32enc() function which handles endianness and
* is also alignment-safe.
*/
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
le32enc(&rfa->link_addr, 0xffffffff);
le32enc(&rfa->rbd_addr, 0xffffffff);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
/* Map the RFA into DMA memory. */
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
error = bus_dmamap_load(sc->fxp_rxmtag, sc->spare_map, rfa,
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
MCLBYTES - RFA_ALIGNMENT_FUDGE, fxp_dma_map_addr,
&rxp->rx_addr, 0);
if (error) {
m_freem(m);
return (error);
}
Always check whether dma map is valid before unloading the map. When fxp(4) intializes Rx buffers for the first time, there is no loaded dma map so attempting to unload it is an invalid operation.
2009-01-08 04:26:44 +00:00
if (rxp->rx_mbuf != NULL)
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_unload(sc->fxp_rxmtag, rxp->rx_map);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
tmp_map = sc->spare_map;
sc->spare_map = rxp->rx_map;
rxp->rx_map = tmp_map;
rxp->rx_mbuf = m;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_rxmtag, rxp->rx_map,
- Clean up the fxp_release() and fxp_detach() functions. - Be sure to teardown the interrupt first so that "kldunload if_fxp" doesn't panic the box. It's now deadlocking rather than crashing, which isn't really better, but I'm unsure this is fxp(4)'s fault. - Change a bus_dmamap_sync() call to also do a BUS_DMASYNC_PREREAD now that we can pass several operations.
2003-04-10 23:15:12 +00:00
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
return (0);
}
static void
fxp_add_rfabuf(struct fxp_softc *sc, struct fxp_rx *rxp)
{
struct fxp_rfa *p_rfa;
struct fxp_rx *p_rx;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
Fix minor bug that would cause the packet statistics updates to stop working under extremely rare circumstances. Also added some additional comments..
1995-12-01 22:41:56 +00:00
/*
* If there are other buffers already on the list, attach this
* one to the end by fixing up the tail to point to this one.
*/
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
if (sc->fxp_desc.rx_head != NULL) {
p_rx = sc->fxp_desc.rx_tail;
p_rfa = (struct fxp_rfa *)
(p_rx->rx_mbuf->m_ext.ext_buf + RFA_ALIGNMENT_FUDGE);
p_rx->rx_next = rxp;
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
le32enc(&p_rfa->link_addr, rxp->rx_addr);
Implemented some optimizations which result in 14 fewer instructions in the receive path.
2000-06-19 00:58:34 +00:00
p_rfa->rfa_control = 0;
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_rxmtag, p_rx->rx_map,
Controller updates RFA via DMA so driver needs synchronization. Add missing BUS_DMASYNC_POSTWRITE and BUS_DMASYNC_PREREAD.
2010-05-14 16:55:13 +00:00
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
} else {
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
rxp->rx_next = NULL;
sc->fxp_desc.rx_head = rxp;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
sc->fxp_desc.rx_tail = rxp;
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
}
static void
fxp_discard_rfabuf(struct fxp_softc *sc, struct fxp_rx *rxp)
{
struct mbuf *m;
struct fxp_rfa *rfa;
m = rxp->rx_mbuf;
m->m_data = m->m_ext.ext_buf;
/*
* Move the data pointer up so that the incoming data packet
* will be 32-bit aligned.
*/
m->m_data += RFA_ALIGNMENT_FUDGE;
/*
* Get a pointer to the base of the mbuf cluster and move
* data start past it.
*/
rfa = mtod(m, struct fxp_rfa *);
m->m_data += sc->rfa_size;
rfa->size = htole16(MCLBYTES - sc->rfa_size - RFA_ALIGNMENT_FUDGE);
rfa->rfa_status = 0;
rfa->rfa_control = htole16(FXP_RFA_CONTROL_EL);
rfa->actual_size = 0;
/*
* Initialize the rest of the RFA. Note that since the RFA
* is misaligned, we cannot store values directly. We're thus
* using the le32enc() function which handles endianness and
* is also alignment-safe.
*/
le32enc(&rfa->link_addr, 0xffffffff);
le32enc(&rfa->rbd_addr, 0xffffffff);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->fxp_rxmtag, rxp->rx_map,
o Introduce a new function, fxp_new_rfabuf which allocates a new Rx buffer and loads DMA map. Also add a function fxp_discard_rfabuf that handles reusing Rx buffer/DMA map. With this change fxp_add_rfabuf just handles appending a new RFA to existing chain. o Initialize mbuf length in fxp_new_rfabuf. o Don't reset rnr and have fxp(4) handle received frames even if it couldn't allocate new Rx buffer. This will make fxp(4) reload updated RFA under rnr case. The rnr would still be reset to 0 if polling is active and fxp(4) processed number of allowed Rx events. o Update if_iqdrops if fxp(4) couldn't allocate Rx buffer. Previously fxp(4) used to try to reuse Rx buffer when new buffer allocation is failed. But fxp(4) didn't take into account loaded DMA map such that the same DMA map was loaded again without unloading the map. There is no reason to unload the loaded map and reload the same map again, just reusing the map is enough. I believe the spare DMA map in softc was introduced to implement this behaviour. Also fxp(4) used to stop Rx processing if once Rx buffer allocation or DMA map load fails which in turn resulted in losing incoming frames under heavy network load. With this change fxp(4) should survive from resource shortage condition.
2008-11-26 05:51:49 +00:00
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Remove the volatile qualifier to apply to fxp_miibus_readreg().
2007-05-30 03:46:04 +00:00
static int
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_miibus_readreg(device_t dev, int phy, int reg)
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
{
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
struct fxp_softc *sc = device_get_softc(dev);
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
int count = 10000;
Added support for newer cards that have the DP83840A PHY chip. Fixed a bug in fxp_mdi_write - a hex number was missing a preceding 0x and this was causing the routine to not wait for a PHY write to complete. Added support for link0, link1, and link2 flags to toggle auto- negotiation, 10/100, and half/full duplex: link0 disable auto-negotiation When set, these flags then have meaning: -link1 10Mbps link1 100Mbps -link2 half duplex link2 full duplex ...needs a manual page.
1997-03-21 08:00:13 +00:00
int value;
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
CSR_WRITE_4(sc, FXP_CSR_MDICONTROL,
(FXP_MDI_READ << 26) | (reg << 16) | (phy << 21));
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
while (((value = CSR_READ_4(sc, FXP_CSR_MDICONTROL)) & 0x10000000) == 0
&& count--)
Added support for newer cards that have the DP83840A PHY chip. Fixed a bug in fxp_mdi_write - a hex number was missing a preceding 0x and this was causing the routine to not wait for a PHY write to complete. Added support for link0, link1, and link2 flags to toggle auto- negotiation, 10/100, and half/full duplex: link0 disable auto-negotiation When set, these flags then have meaning: -link1 10Mbps link1 100Mbps -link2 half duplex link2 full duplex ...needs a manual page.
1997-03-21 08:00:13 +00:00
DELAY(10);
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
if (count <= 0)
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
device_printf(dev, "fxp_miibus_readreg: timed out\n");
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
Added support for newer cards that have the DP83840A PHY chip. Fixed a bug in fxp_mdi_write - a hex number was missing a preceding 0x and this was causing the routine to not wait for a PHY write to complete. Added support for link0, link1, and link2 flags to toggle auto- negotiation, 10/100, and half/full duplex: link0 disable auto-negotiation When set, these flags then have meaning: -link1 10Mbps link1 100Mbps -link2 half duplex link2 full duplex ...needs a manual page.
1997-03-21 08:00:13 +00:00
return (value & 0xffff);
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
}
writereg returns an int.
2009-02-05 19:33:35 +00:00
static int
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_miibus_writereg(device_t dev, int phy, int reg, int value)
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
{
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
struct fxp_softc *sc = device_get_softc(dev);
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
int count = 10000;
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
CSR_WRITE_4(sc, FXP_CSR_MDICONTROL,
(FXP_MDI_WRITE << 26) | (reg << 16) | (phy << 21) |
(value & 0xffff));
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
while ((CSR_READ_4(sc, FXP_CSR_MDICONTROL) & 0x10000000) == 0 &&
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
count--)
Added support for newer cards that have the DP83840A PHY chip. Fixed a bug in fxp_mdi_write - a hex number was missing a preceding 0x and this was causing the routine to not wait for a PHY write to complete. Added support for link0, link1, and link2 flags to toggle auto- negotiation, 10/100, and half/full duplex: link0 disable auto-negotiation When set, these flags then have meaning: -link1 10Mbps link1 100Mbps -link2 half duplex link2 full duplex ...needs a manual page.
1997-03-21 08:00:13 +00:00
DELAY(10);
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
if (count <= 0)
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
device_printf(dev, "fxp_miibus_writereg: timed out\n");
writereg returns an int.
2009-02-05 19:33:35 +00:00
return (0);
Fixed two deficiencies in the driver that have existed since it was written: 1) Full duplex mode is now supported (and works!) 2) The 10Mbps-only PCI Pro/10 should now work (untested, however) Thanks to Justin Gibbs for providing a PCI bus analyzer trace while the Intel Windows driver was configuring the board...this made it possible to figure out the mystery bit that I wasn't setting in the PHY for full duplex to work.
1997-03-17 11:08:16 +00:00
}
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
static int
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
{
Clean up Ethernet drivers: - fill in and use ifp->if_softc - use if_bpf rather than private cookie variables - change bpf interface to take advantage of this - call ether_ifattach() directly from Ethernet drivers - delete kludge in if_attach() that did this indirectly
1996-02-06 18:51:28 +00:00
struct fxp_softc *sc = ifp->if_softc;
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
struct ifreq *ifr = (struct ifreq *)data;
Fix a whitespace bogon. Pointed out by: ps
2001-03-12 21:42:45 +00:00
struct mii_data *mii;
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
int flag, mask, error = 0, reinit;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
switch (command) {
case SIOCSIFFLAGS:
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
FXP_LOCK(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
* If interface is marked up and not running, then start it.
* If it is marked down and running, stop it.
* XXX If it's up then re-initialize it. This is so flags
* such as IFF_PROMISC are handled.
*/
if (ifp->if_flags & IFF_UP) {
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
if (((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0) &&
((ifp->if_flags ^ sc->if_flags) &
(IFF_PROMISC | IFF_ALLMULTI | IFF_LINK0)) != 0)
fxp_init_body(sc);
else if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
fxp_init_body(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
} else {
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
all: Removed ifnet.if_init and ifnet.if_reset as they are generally unused. Change the parameter passed to if_watchdog to be a ifnet * rather than a unit number. All of this is an attempt to move toward not needing an array of softc pointers (which is usually static in size) to point to the driver softc. if_ed.c: Changed some of the argument passing to some functions to make a little more sense. if_ep.c, if_vx.c: Killed completely bogus use of if_timer. It was being set in such a way that the interface was being reset once per second (blech!).
1995-12-05 02:01:59 +00:00
fxp_stop(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
sc->if_flags = ifp->if_flags;
- Use callout_init_mtx() to close a small race between callout_stop() and the timeout routine. - Fix locking in detach. - Add locking in shutdown. - Don't mess with the PCI command register in resume, the PCI bus driver already does this for us. - Add locking to the non-serial ifmedia routines. - Fix locking in ioctl. - Remove spls and support for 4.x. MFC after: 1 week
2005-08-08 21:03:54 +00:00
FXP_UNLOCK(sc);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != 0)
fxp_init(sc);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
break;
case SIOCSIFMEDIA:
case SIOCGIFMEDIA:
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
if (sc->miibus != NULL) {
mii = device_get_softc(sc->miibus);
error = ifmedia_ioctl(ifp, ifr,
&mii->mii_media, command);
} else {
error = ifmedia_ioctl(ifp, ifr, &sc->sc_media, command);
}
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
break;
Implemented per-interface polling(4) control.
2004-04-11 15:35:49 +00:00
case SIOCSIFCAP:
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
reinit = 0;
Teach fxp(4) to control VLAN_MTU in the hardware. Now reception of extended frames can be toggled through ioctl(SIOCSIFCAP). The card will also receive extended frames when in promiscuous mode.
2004-05-25 14:49:46 +00:00
mask = ifp->if_capenable ^ ifr->ifr_reqcap;
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(fxp_poll, ifp);
if (error)
return(error);
FXP_LOCK(sc);
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL,
FXP_SCB_INTR_DISABLE);
ifp->if_capenable |= IFCAP_POLLING;
FXP_UNLOCK(sc);
} else {
error = ether_poll_deregister(ifp);
/* Enable interrupts in any case */
FXP_LOCK(sc);
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, 0);
ifp->if_capenable &= ~IFCAP_POLLING;
FXP_UNLOCK(sc);
}
}
#endif
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
FXP_LOCK(sc);
if ((mask & IFCAP_TXCSUM) != 0 &&
(ifp->if_capabilities & IFCAP_TXCSUM) != 0) {
ifp->if_capenable ^= IFCAP_TXCSUM;
if ((ifp->if_capenable & IFCAP_TXCSUM) != 0)
ifp->if_hwassist |= FXP_CSUM_FEATURES;
else
ifp->if_hwassist &= ~FXP_CSUM_FEATURES;
}
if ((mask & IFCAP_RXCSUM) != 0 &&
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
(ifp->if_capabilities & IFCAP_RXCSUM) != 0) {
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
ifp->if_capenable ^= IFCAP_RXCSUM;
Implement Rx checksum offload for 82559 or later controllers. 82559 or later controllers added simple checksum calculation logic in RU. For backward compatibility the computed checksum is appended at the end of the data posted to Rx buffer. This type of simple checksum calculation support had been used on several vendors such as Sun HME/GEM, SysKonnect GENESIS and Marvell Yukon controllers. Because this type of checksum offload support requires parsing of received frame and pseudo checksum calculation with software routine it still consumes more CPU cycles than that of full-fledged checksum offload controller. But it's still better than software checksum calculation.
2008-11-26 06:36:53 +00:00
if ((sc->flags & FXP_FLAG_82559_RXCSUM) != 0)
reinit++;
}
Implement TSO for 82550/82551 controllers. o Configure controller to use dynamic TBD as TSO requires that operation mode. o Add a dummy TBD to tx_cb_u as TSO can access one more TBD in TSO operation. o Increase a DMA segment size to 4096 to hold a full IP segment with link layer header. o Unlike other TSO capable controllers, 82550/82551 does not modify the first IP packet in TSO operation so driver should create an IP packet with proper header. Subsequent IP packets are generated from the header information in the first IP packet header. Likewise pseudo checksum also should be computed by driver for the first packet. o TSO requires one more TBD to hold total TCP payload. To make code simple for TSO/non-TSO case, increase the index of the first available TBD array. o Remove KASSERT that checks the size of a DMA segment should be less than or equal to MCLBYTES as it's no longer valid in TSO. o Tx threshold and number of TBDs field is used to store MSS in TSO. So don't set the Tx threshold in TSO case.
2008-11-26 07:36:17 +00:00
if ((mask & IFCAP_TSO4) != 0 &&
(ifp->if_capabilities & IFCAP_TSO4) != 0) {
ifp->if_capenable ^= IFCAP_TSO4;
if ((ifp->if_capenable & IFCAP_TSO4) != 0)
ifp->if_hwassist |= CSUM_TSO;
else
ifp->if_hwassist &= ~CSUM_TSO;
}
Add basic WOL support for 82550/82551/82558 and 82559 based controllers. ICH based controllers are treated as 82559. 82557, earlier revision of 82558 and 82559ER have no WOL capability. o WOL support requires help of a firmware so add check whether hardware is capable of handling magic frames by reading EEPROM. o Enable accepting WOL frames only when hardware is about to suspend or shutdown. Previously fxp(4) used to allow receipt of magic frame under normal operation mode which could cause hardware hang if magic frame is received by hardware. Datasheet clearly states driver should not allow WOL frames under normal operation mode. o Disable WOL frame reception in device attach so have fxp(4) immunize against system hang which can be triggered by magic packets when the hardware is not in fully initialized state. o Don't reset all hardware configuration data in fxp_stop() otherwise important configuration data is lost and this would reset WOL configuration to default state which in turn cause hardware hang on receipt of magic frames. To fix the issue, preserve hardware configuration data by issuing a selective reset. o Explicitly disable interrupts after issuing selective reset as reset may unmask interrupts. Tested by: Alexey Shuvaev < shuvaev <> physik DOT uni-wuerzburg DOT de >
2008-11-27 01:57:23 +00:00
if ((mask & IFCAP_WOL_MAGIC) != 0 &&
(ifp->if_capabilities & IFCAP_WOL_MAGIC) != 0)
ifp->if_capenable ^= IFCAP_WOL_MAGIC;
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
if ((mask & IFCAP_VLAN_MTU) != 0 &&
(ifp->if_capabilities & IFCAP_VLAN_MTU) != 0) {
Teach fxp(4) to control VLAN_MTU in the hardware. Now reception of extended frames can be toggled through ioctl(SIOCSIFCAP). The card will also receive extended frames when in promiscuous mode.
2004-05-25 14:49:46 +00:00
ifp->if_capenable ^= IFCAP_VLAN_MTU;
if (sc->revision != FXP_REV_82557)
flag = FXP_FLAG_LONG_PKT_EN;
else /* a hack to get long frames on the old chip */
flag = FXP_FLAG_SAVE_BAD;
sc->flags ^= flag;
if (ifp->if_flags & IFF_UP)
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
reinit++;
Teach fxp(4) to control VLAN_MTU in the hardware. Now reception of extended frames can be toggled through ioctl(SIOCSIFCAP). The card will also receive extended frames when in promiscuous mode.
2004-05-25 14:49:46 +00:00
}
Add VLAN hardware tag insertion/stripping support. Tx/Rx checksum offload for VLAN frames are also supported. The VLAN hardware assistance is available only on 82550/82551 based controllers. While I'm here change the confusing name of bit1 in byte 22 of configuration block to vlan_drop_en. The bit controls whether hardware strips VLAN tagged frame or not. Special thanks to wpaul who sent valuable VLAN related information to me. Tested on: i386, sparc64
2008-12-02 02:30:12 +00:00
if ((mask & IFCAP_VLAN_HWTAGGING) != 0 &&
(ifp->if_capabilities & IFCAP_VLAN_HWTAGGING) != 0) {
ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
reinit++;
}
if (reinit > 0 && ifp->if_flags & IFF_UP)
Fix Tx/Rx checksum offload ioctl handling and make Rx handler honor checksum offload configuration. Now checksum offload can be controlled by ifconfig(8). While I'm here add an additional check for interface capabilities before applying user's request.
2008-11-25 01:33:16 +00:00
fxp_init_body(sc);
FXP_UNLOCK(sc);
Add VLAN hardware tag insertion/stripping support. Tx/Rx checksum offload for VLAN frames are also supported. The VLAN hardware assistance is available only on 82550/82551 based controllers. While I'm here change the confusing name of bit1 in byte 22 of configuration block to vlan_drop_en. The bit controls whether hardware strips VLAN tagged frame or not. Special thanks to wpaul who sent valuable VLAN related information to me. Tested on: i386, sparc64
2008-12-02 02:30:12 +00:00
VLAN_CAPABILITIES(ifp);
Implemented per-interface polling(4) control.
2004-04-11 15:35:49 +00:00
break;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
default:
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
error = ether_ioctl(ifp, command, data);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
}
return (error);
}
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
/*
* Fill in the multicast address list and return number of entries.
*/
static int
fxp_mc_addrs(struct fxp_softc *sc)
{
struct fxp_cb_mcs *mcsp = sc->mcsp;
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
struct ifnet *ifp = sc->ifp;
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
struct ifmultiaddr *ifma;
int nmcasts;
nmcasts = 0;
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
if ((ifp->if_flags & IFF_ALLMULTI) == 0) {
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_rlock(ifp);
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
if (nmcasts >= MAXMCADDR) {
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
ifp->if_flags |= IFF_ALLMULTI;
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
nmcasts = 0;
break;
}
bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
- Endian fixes. - 6 -> ETHER_ADDR_LEN.
2003-04-08 17:21:15 +00:00
&sc->mcsp->mc_addr[nmcasts][0], ETHER_ADDR_LEN);
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
nmcasts++;
}
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_runlock(ifp);
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
}
- Endian fixes. - 6 -> ETHER_ADDR_LEN.
2003-04-08 17:21:15 +00:00
mcsp->mc_cnt = htole16(nmcasts * ETHER_ADDR_LEN);
Explicitly reload the multicast filters when the hardware is reinitialized instead of relying on the previous filters to be present. Back out r1.125, as a reset is needed to unload any existing microcode, (which clears the multicast addresses), as it is superceded by this change.
2002-01-07 15:08:54 +00:00
return (nmcasts);
}
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
/*
* Program the multicast filter.
*
* We have an artificial restriction that the multicast setup command
* must be the first command in the chain, so we take steps to ensure
Implemented a more sophisticated mechanism for handling transmitter interrupts which now defers them until the transmit queue if filled up with completed buffers. This has two advantages: first, it reduces the number of transmitter interrupts to just 1/120th of the rate that they occured previously, and two, running down many buffers at once has much improved cache effects.
1998-08-04 08:53:12 +00:00
* this. By requiring this, it allows us to keep up the performance of
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
* the pre-initialized command ring (esp. link pointers) by not actually
Seventy-odd "its" / "it's" typos in comments fixed as per kern/6108.
1998-04-17 22:37:19 +00:00
* inserting the mcsetup command in the ring - i.e. its link pointer
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
* points to the TxCB ring, but the mcsetup descriptor itself is not part
* of it. We then can do 'CU_START' on the mcsetup descriptor and have it
* lead into the regular TxCB ring when it completes.
*/
static void
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
fxp_mc_setup(struct fxp_softc *sc)
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
{
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
struct fxp_cb_mcs *mcsp;
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
int count;
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
Abstract the locking in fxp(4) a bit more by using macros for mtx_assert() and mtx_owned(), as it is done in other places, for instance proc locking.
2004-06-02 22:59:57 +00:00
FXP_LOCK_ASSERT(sc, MA_OWNED);
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
mcsp = sc->mcsp;
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
mcsp->cb_status = 0;
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
mcsp->cb_command = htole16(FXP_CB_COMMAND_MCAS | FXP_CB_COMMAND_EL);
mcsp->link_addr = 0xffffffff;
fxp_mc_addrs(sc);
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
/*
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
* Wait until command unit is idle. This should never be the
* case when nothing is queued, but make sure anyway.
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
*/
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
count = 100;
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
while ((CSR_READ_1(sc, FXP_CSR_SCB_RUSCUS) >> 6) !=
FXP_SCB_CUS_IDLE && --count)
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
DELAY(10);
if (count == 0) {
Convert the fxp driver to miibus, which involves ripping out the PHY logic and media bits. Support for Intel PHYs can now be found in dev/mii/inphy.c. Clean up the driver, and add various 82558 and 82559 specific bits.
2001-03-12 21:30:52 +00:00
device_printf(sc->dev, "command queue timeout\n");
Added support for APM suspend/resume. PR: 18756 Submitted by: mike ryan <msr@elision.org>, with modifications by me.
2000-09-17 22:12:12 +00:00
return;
}
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
/*
* Start the multicast setup command.
*/
fxp_scb_wait(sc);
Introduce Rx mbuf dma tag and use it in Rx path. Previously it used common mbuf dma tag for both Tx and Rx path but Rx buffer should have single DMA segment and maximum buffer size of the segment should be less than MCLBYTES. fxp(4) also have to check Tx completion status which was updated by DMA so we need BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTWRITE synchronization in Tx path. Fix all misuse of bus_dmamap_sync(9) in fxp(4). I guess this change shall fix occasional driver breakage in PAE environments. While I'm here add error messages of dma tag/buffer creation and correct messages.
2009-06-21 06:06:43 +00:00
bus_dmamap_sync(sc->mcs_tag, sc->mcs_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->mcs_addr);
Add workaround for embedded NICs, in particular, the 815E boards. There appears to be a bug where the chip will lock up when running in 10Mb/s mode.
2001-05-17 23:50:24 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);
Overhaul fxp(4) multicast filter programming. fxp(4) hardwares do not allow multicast filter programming when controller is busy to send/receive frames. So it used to mark need_mcsetup bit and defer multicast filter programming until controller becomes idle state. To detect when the controller is idle fxp(4) relied on Tx completion interrupt with NOP command and fxp_start_body and fxp_intr_body had to see whether pending multicast filter programming was requested. This resulted in very complex logic and sometimes it did not work as expected. Since the controller should be in idle state before any multicast filter modifications I changed it to reinitialize the controller whenever multicast filter programming is required. This is the same way what OpenBSD and NetBSD does. Also I added IFF_DRV_RUNNING check in ioctl handler so controller would be reinitialized only if it is absolutely needed. With this change I guess we can remove fxp(4) DELAY hack in ifioctl for IPv6 case.
2009-06-21 07:17:49 +00:00
/* ...and wait for it to complete. */
fxp_dma_wait(sc, &mcsp->cb_status, sc->mcs_tag, sc->mcs_map);
Work around a bug in the 82557 NIC where the receiver will lock up if it is in 10Mbps mode and gets certain types of garbage prior to the packet header. The work-around involves reprogramming the multicast filter if nothing is received in some number of seconds (currently set at 15). As a side effect, implemented complete support for multicasting rather than the previous 'receive all multicasts' hack, since we now have the ability to program the filter table. Fixed a serious bug which crept in with the timeout() changes; the cookie was only saved on the first timeout() call in fxp_init() and wasn't updated in the most common place in fxp_stats_update() when the timeout was rescheduled. This bug would have resulted in an eventual panic if fxp_stop() was called (which happens when any interface flags are changed, for example). Fixed a bug in Alpha support that would have caused the TxCB descriptor chain to span a page boundry, causing serious problems if the pages didn't happen to be contiguous. Removed some gratuitous bit masking that was left over from an older implementation. Fixed a bug where too much was copied from the configuration template, spilling over into memory that followed it. Fixed handling of if_timer...it was cleared too early in some cases.
1997-09-29 11:27:43 +00:00
}
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
static uint32_t fxp_ucode_d101a[] = D101_A_RCVBUNDLE_UCODE;
static uint32_t fxp_ucode_d101b0[] = D101_B0_RCVBUNDLE_UCODE;
static uint32_t fxp_ucode_d101ma[] = D101M_B_RCVBUNDLE_UCODE;
static uint32_t fxp_ucode_d101s[] = D101S_RCVBUNDLE_UCODE;
static uint32_t fxp_ucode_d102[] = D102_B_RCVBUNDLE_UCODE;
static uint32_t fxp_ucode_d102c[] = D102_C_RCVBUNDLE_UCODE;
Add a microcode to implement receive bundling for 82551 chipsets with a revision ID of 0x0f (D102 E-step). MFC after: 2 weeks Tested by: pav
2005-04-21 19:34:57 +00:00
static uint32_t fxp_ucode_d102e[] = D102_E_RCVBUNDLE_UCODE;
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
#define UCODE(x) x, sizeof(x)/sizeof(uint32_t)
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
struct ucode {
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
uint32_t revision;
uint32_t *ucode;
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
int length;
u_short int_delay_offset;
u_short bundle_max_offset;
} ucode_table[] = {
{ FXP_REV_82558_A4, UCODE(fxp_ucode_d101a), D101_CPUSAVER_DWORD, 0 },
{ FXP_REV_82558_B0, UCODE(fxp_ucode_d101b0), D101_CPUSAVER_DWORD, 0 },
{ FXP_REV_82559_A0, UCODE(fxp_ucode_d101ma),
D101M_CPUSAVER_DWORD, D101M_CPUSAVER_BUNDLE_MAX_DWORD },
{ FXP_REV_82559S_A, UCODE(fxp_ucode_d101s),
D101S_CPUSAVER_DWORD, D101S_CPUSAVER_BUNDLE_MAX_DWORD },
{ FXP_REV_82550, UCODE(fxp_ucode_d102),
D102_B_CPUSAVER_DWORD, D102_B_CPUSAVER_BUNDLE_MAX_DWORD },
{ FXP_REV_82550_C, UCODE(fxp_ucode_d102c),
D102_C_CPUSAVER_DWORD, D102_C_CPUSAVER_BUNDLE_MAX_DWORD },
Be more conservative when enabling extended features. There are fxp(4) NICs out there that have an utterly bogus revision ID. Reported by: Denis Shaposhnikov <dsh@vlink.ru>
2005-04-22 13:05:53 +00:00
{ FXP_REV_82551_F, UCODE(fxp_ucode_d102e),
Add a microcode to implement receive bundling for 82551 chipsets with a revision ID of 0x0f (D102 E-step). MFC after: 2 weeks Tested by: pav
2005-04-21 19:34:57 +00:00
D102_E_CPUSAVER_DWORD, D102_E_CPUSAVER_BUNDLE_MAX_DWORD },
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
{ 0, NULL, 0, 0, 0 }
};
static void
fxp_load_ucode(struct fxp_softc *sc)
{
struct ucode *uc;
struct fxp_cb_ucode *cbp;
Make interrupt coalescing work on big endian systems. Also change struct ucode.length to be in number of elements (u_int32_t) to help endian handling. MFC after: 2 weeks
2004-11-17 04:25:10 +00:00
int i;
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
for (uc = ucode_table; uc->ucode != NULL; uc++)
if (sc->revision == uc->revision)
break;
if (uc->ucode == NULL)
return;
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
cbp = (struct fxp_cb_ucode *)sc->fxp_desc.cbl_list;
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
cbp->cb_status = 0;
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
cbp->cb_command = htole16(FXP_CB_COMMAND_UCODE | FXP_CB_COMMAND_EL);
cbp->link_addr = 0xffffffff; /* (no) next command */
Make interrupt coalescing work on big endian systems. Also change struct ucode.length to be in number of elements (u_int32_t) to help endian handling. MFC after: 2 weeks
2004-11-17 04:25:10 +00:00
for (i = 0; i < uc->length; i++)
cbp->ucode[i] = htole32(uc->ucode[i]);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
if (uc->int_delay_offset)
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
*(uint16_t *)&cbp->ucode[uc->int_delay_offset] =
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
htole16(sc->tunable_int_delay + sc->tunable_int_delay / 2);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
if (uc->bundle_max_offset)
Cleanup: u_intXX_t -> uintX_t conversion.
2005-03-06 05:07:26 +00:00
*(uint16_t *)&cbp->ucode[uc->bundle_max_offset] =
- Instead of rolling our own alignment-safe function, use le32enc() which deals with both endianness and alignment issues. - Collect low-hanging fruits for endianness safety. - Use 0xffffffff instead of -1 where appropriate.
2003-04-06 01:04:17 +00:00
htole16(sc->tunable_bundle_max);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
/*
* Download the ucode to the chip.
*/
fxp_scb_wait(sc);
Controller will dma SCB command status for a given command and driver should read updated status back after issuing a SCB command. To send a command to controller and read updated status back, driver should synchronize both memory read and write operations with device. Fix bus_dmamap_sync operation specifier used in fxp_dma_wait() by adding both memory read and memory write operations.
2009-06-10 01:15:30 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Convert the fxp(4) driver to the busdma API. This patch is rather big because I had to significantly redesign the driver to make the busdma conversion possible. Most notably, hardware and software structures were carefully splitted to get rid of all the structs overlapping evilness. Special thanks to phk and Richard Puga <puga@mauibuilt.com> for providing me with fxp(4) hardware to do this work. Thanks to marcel for testing this on ia64, and to Fred Clift <fclift@verio.net> for testing this on alpha. Tested on: i386, ia64, alpha
2003-04-02 16:47:16 +00:00
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->fxp_desc.cbl_addr);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
fxp_scb_cmd(sc, FXP_SCB_COMMAND_CU_START);
/* ...and wait for it to complete. */
- Pass a busdma tag and a busdma map to fxp_dma_wait() so that it can do the necessary bus_dmamap_sync() calls. - While I'm here, reorder arguments to match other functions.
2003-04-08 18:56:45 +00:00
fxp_dma_wait(sc, &cbp->cb_status, sc->cbl_tag, sc->cbl_map);
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
device_printf(sc->dev,
"Microcode loaded, int_delay: %d usec bundle_max: %d\n",
Whitespace fix.
2008-11-25 00:48:05 +00:00
sc->tunable_int_delay,
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
uc->bundle_max_offset == 0 ? 0 : sc->tunable_bundle_max);
sc->flags |= FXP_FLAG_UCODE;
}
Export hardware MAC statistics through sysctl node. Previously fxp(4) already used to extract most hardware MAC statistics but it didn't show them. With this change, all MAC statistics counters are exported. Because there are a couple of new counters for 82558 and 82559, enable extended MAC statistics functionality to get these counters. Accoring to public data sheet, 82559 MAC statistics return 24 DWORD counters(3 counters are unknown at this moment) so increase MAC counter structure to meet the MAC statistics block size. The completion of MAC counter dump is now checked against FXP_STATS_DR_COMPLETE status code which is appended at the end of status block. Previously fxp(4) ignored the status of the FXP_SCB_COMMAND_CU_DUMPRESET command. fxp(4) does not wait for the completion of pending command before issuing FXP_SCB_COMMAND_CU_DUMPRESET. Instead it skips the command and try it next time. This scheme may show better performance but there is chance to loose updated counters after stopping controller. So make sure to update MAC statistics in fxp_stop(). While I'm here move sysctl node creation to fxp_sysctl_node(). Tested by: Larry Baird < lab <> gta dot com >
2010-05-09 22:16:15 +00:00
#define FXP_SYSCTL_STAT_ADD(c, h, n, p, d) \
SYSCTL_ADD_UINT(c, h, OID_AUTO, n, CTLFLAG_RD, p, 0, d)
static void
fxp_sysctl_node(struct fxp_softc *sc)
{
struct sysctl_ctx_list *ctx;
struct sysctl_oid_list *child, *parent;
struct sysctl_oid *tree;
struct fxp_hwstats *hsp;
ctx = device_get_sysctl_ctx(sc->dev);
child = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->dev));
SYSCTL_ADD_PROC(ctx, child,
OID_AUTO, "int_delay", CTLTYPE_INT | CTLFLAG_RW,
&sc->tunable_int_delay, 0, sysctl_hw_fxp_int_delay, "I",
"FXP driver receive interrupt microcode bundling delay");
SYSCTL_ADD_PROC(ctx, child,
OID_AUTO, "bundle_max", CTLTYPE_INT | CTLFLAG_RW,
&sc->tunable_bundle_max, 0, sysctl_hw_fxp_bundle_max, "I",
"FXP driver receive interrupt microcode bundle size limit");
SYSCTL_ADD_INT(ctx, child,OID_AUTO, "rnr", CTLFLAG_RD, &sc->rnr, 0,
"FXP RNR events");
SYSCTL_ADD_INT(ctx, child,
OID_AUTO, "noflow", CTLFLAG_RW, &sc->tunable_noflow, 0,
"FXP flow control disabled");
/*
* Pull in device tunables.
*/
sc->tunable_int_delay = TUNABLE_INT_DELAY;
sc->tunable_bundle_max = TUNABLE_BUNDLE_MAX;
sc->tunable_noflow = 1;
(void) resource_int_value(device_get_name(sc->dev),
device_get_unit(sc->dev), "int_delay", &sc->tunable_int_delay);
(void) resource_int_value(device_get_name(sc->dev),
device_get_unit(sc->dev), "bundle_max", &sc->tunable_bundle_max);
(void) resource_int_value(device_get_name(sc->dev),
device_get_unit(sc->dev), "noflow", &sc->tunable_noflow);
sc->rnr = 0;
hsp = &sc->fxp_hwstats;
tree = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "stats", CTLFLAG_RD,
NULL, "FXP statistics");
parent = SYSCTL_CHILDREN(tree);
/* Rx MAC statistics. */
tree = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO, "rx", CTLFLAG_RD,
NULL, "Rx MAC statistics");
child = SYSCTL_CHILDREN(tree);
FXP_SYSCTL_STAT_ADD(ctx, child, "good_frames",
&hsp->rx_good, "Good frames");
FXP_SYSCTL_STAT_ADD(ctx, child, "crc_errors",
&hsp->rx_crc_errors, "CRC errors");
FXP_SYSCTL_STAT_ADD(ctx, child, "alignment_errors",
&hsp->rx_alignment_errors, "Alignment errors");
FXP_SYSCTL_STAT_ADD(ctx, child, "rnr_errors",
&hsp->rx_rnr_errors, "RNR errors");
FXP_SYSCTL_STAT_ADD(ctx, child, "overrun_errors",
&hsp->rx_overrun_errors, "Overrun errors");
FXP_SYSCTL_STAT_ADD(ctx, child, "cdt_errors",
&hsp->rx_cdt_errors, "Collision detect errors");
FXP_SYSCTL_STAT_ADD(ctx, child, "shortframes",
&hsp->rx_shortframes, "Short frame errors");
if (sc->revision >= FXP_REV_82558_A4) {
FXP_SYSCTL_STAT_ADD(ctx, child, "pause",
&hsp->rx_pause, "Pause frames");
FXP_SYSCTL_STAT_ADD(ctx, child, "controls",
&hsp->rx_controls, "Unsupported control frames");
}
if (sc->revision >= FXP_REV_82559_A0)
FXP_SYSCTL_STAT_ADD(ctx, child, "tco",
&hsp->rx_tco, "TCO frames");
/* Tx MAC statistics. */
tree = SYSCTL_ADD_NODE(ctx, parent, OID_AUTO, "tx", CTLFLAG_RD,
NULL, "Tx MAC statistics");
child = SYSCTL_CHILDREN(tree);
FXP_SYSCTL_STAT_ADD(ctx, child, "good_frames",
&hsp->tx_good, "Good frames");
FXP_SYSCTL_STAT_ADD(ctx, child, "maxcols",
&hsp->tx_maxcols, "Maximum collisions errors");
FXP_SYSCTL_STAT_ADD(ctx, child, "latecols",
&hsp->tx_latecols, "Late collisions errors");
FXP_SYSCTL_STAT_ADD(ctx, child, "underruns",
&hsp->tx_underruns, "Underrun errors");
FXP_SYSCTL_STAT_ADD(ctx, child, "lostcrs",
&hsp->tx_lostcrs, "Lost carrier sense");
FXP_SYSCTL_STAT_ADD(ctx, child, "deffered",
&hsp->tx_deffered, "Deferred");
FXP_SYSCTL_STAT_ADD(ctx, child, "single_collisions",
&hsp->tx_single_collisions, "Single collisions");
FXP_SYSCTL_STAT_ADD(ctx, child, "multiple_collisions",
&hsp->tx_multiple_collisions, "Multiple collisions");
FXP_SYSCTL_STAT_ADD(ctx, child, "total_collisions",
&hsp->tx_total_collisions, "Total collisions");
if (sc->revision >= FXP_REV_82558_A4)
FXP_SYSCTL_STAT_ADD(ctx, child, "pause",
&hsp->tx_pause, "Pause frames");
if (sc->revision >= FXP_REV_82559_A0)
FXP_SYSCTL_STAT_ADD(ctx, child, "tco",
&hsp->tx_tco, "TCO frames");
}
#undef FXP_SYSCTL_STAT_ADD
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
static int
sysctl_int_range(SYSCTL_HANDLER_ARGS, int low, int high)
{
int error, value;
value = *(int *)arg1;
error = sysctl_handle_int(oidp, &value, 0, req);
if (error || !req->newptr)
return (error);
if (value < low || value > high)
return (EINVAL);
*(int *)arg1 = value;
return (0);
}
/*
* Interrupt delay is expressed in microseconds, a multiplier is used
Whitespace fix.
2008-11-25 00:48:05 +00:00
* to convert this to the appropriate clock ticks before using.
Add support for loadable microcode which implements interrupt coalescing and packet bundling. Make the microcode settings controllable via sysctl and loader tunables. Submitted by: Marko Zec <zec@tel.fer.hr> (with some munging and dynamic sysctl support by me) Also extend the workaround for Dynamic Standby mode to later '559 chips, not just the ICH2 variants.
2001-10-25 05:32:01 +00:00
*/
static int
sysctl_hw_fxp_int_delay(SYSCTL_HANDLER_ARGS)
{
return (sysctl_int_range(oidp, arg1, arg2, req, 300, 3000));
}
static int
sysctl_hw_fxp_bundle_max(SYSCTL_HANDLER_ARGS)
{
return (sysctl_int_range(oidp, arg1, arg2, req, 1, 0xffff));
}
Reference in New Issue Copy Permalink