d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
eff70bec75
freebsd-dev/sys/dev/fxp/if_fxp.c

2769 lines
78 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
*/
#include <sys/param.h>
#include <sys/systm.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/mbuf.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
/* #include <sys/mutex.h> */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
#include <sys/kernel.h>
Add missing <sys/module.h> includes
2004-05-30 20:08:47 +00:00
#include <sys/module.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>
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
#include <net/if.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>
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
#include <net/bpf.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 <sys/sockio.h>
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
#include <sys/bus.h>
#include <machine/bus.h>
#include <sys/rman.h>
#include <machine/resource.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
Make INET a proper option. This will not make any of object files that LINT create change; there might be differences with INET disabled, but hardly anything compiled before without INET anyway. Now the 'obvious' things will give a proper error if compiled without inet - ipx_ip, ipfw, tcp_debug. The only thing that _should_ work (but can't be made to compile reasonably easily) is sppp :-( This commit move struct arpcom from <netinet/if_ether.h> to <net/if_arp.h>.
1998-01-08 23:42:31 +00:00
#include <net/ethernet.h>
#include <net/if_arp.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
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 <machine/clock.h> /* for DELAY */
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +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
#include <net/if_types.h>
#include <net/if_vlan_var.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
#ifdef FXP_IP_CSUM_WAR
#include <netinet/in.h>
#include <netinet/in_systm.h>
#include <netinet/ip.h>
#include <machine/in_cksum.h>
#endif
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
* must be one or must be zero. Set up a template for these bits
* only, (assuming a 82557 chip) leaving the actual configuration
* to fxp_init.
*
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 */
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
0x5 /* 21 */
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_ident {
u_int16_t devid;
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
int16_t revid; /* -1 matches anything */
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
char *name;
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
/*
* 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[] = {
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
{ 0x1029, -1, "Intel 82559 PCI/CardBus Pro/100" },
{ 0x1030, -1, "Intel 82559 Pro/100 Ethernet" },
{ 0x1031, -1, "Intel 82801CAM (ICH3) Pro/100 VE Ethernet" },
{ 0x1032, -1, "Intel 82801CAM (ICH3) Pro/100 VE Ethernet" },
{ 0x1033, -1, "Intel 82801CAM (ICH3) Pro/100 VM Ethernet" },
{ 0x1034, -1, "Intel 82801CAM (ICH3) Pro/100 VM Ethernet" },
{ 0x1035, -1, "Intel 82801CAM (ICH3) Pro/100 Ethernet" },
{ 0x1036, -1, "Intel 82801CAM (ICH3) Pro/100 Ethernet" },
{ 0x1037, -1, "Intel 82801CAM (ICH3) Pro/100 Ethernet" },
{ 0x1038, -1, "Intel 82801CAM (ICH3) Pro/100 VM Ethernet" },
{ 0x1039, -1, "Intel 82801DB (ICH4) Pro/100 VE Ethernet" },
{ 0x103A, -1, "Intel 82801DB (ICH4) Pro/100 Ethernet" },
{ 0x103B, -1, "Intel 82801DB (ICH4) Pro/100 VM Ethernet" },
{ 0x103C, -1, "Intel 82801DB (ICH4) Pro/100 Ethernet" },
{ 0x103D, -1, "Intel 82801DB (ICH4) Pro/100 VE Ethernet" },
{ 0x103E, -1, "Intel 82801DB (ICH4) Pro/100 VM Ethernet" },
{ 0x1050, -1, "Intel 82801BA (D865) Pro/100 VE Ethernet" },
Add new ID for Intel 82562ET (ICH5/ICH5R) Pro/100 VE Ethernet. Submitted by: Stefan Bethke PR: 61320
2004-04-07 15:47:14 +00:00
{ 0x1051, -1, "Intel 82562ET (ICH5/ICH5R) Pro/100 VE Ethernet" },
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
{ 0x1059, -1, "Intel 82551QM Pro/100 M Mobile Connection" },
{ 0x1209, -1, "Intel 82559ER Embedded 10/100 Ethernet" },
{ 0x1229, 0x01, "Intel 82557 Pro/100 Ethernet" },
{ 0x1229, 0x02, "Intel 82557 Pro/100 Ethernet" },
{ 0x1229, 0x03, "Intel 82557 Pro/100 Ethernet" },
{ 0x1229, 0x04, "Intel 82558 Pro/100 Ethernet" },
{ 0x1229, 0x05, "Intel 82558 Pro/100 Ethernet" },
{ 0x1229, 0x06, "Intel 82559 Pro/100 Ethernet" },
{ 0x1229, 0x07, "Intel 82559 Pro/100 Ethernet" },
{ 0x1229, 0x08, "Intel 82559 Pro/100 Ethernet" },
{ 0x1229, 0x09, "Intel 82559ER Pro/100 Ethernet" },
{ 0x1229, 0x0c, "Intel 82550 Pro/100 Ethernet" },
{ 0x1229, 0x0d, "Intel 82550 Pro/100 Ethernet" },
{ 0x1229, 0x0e, "Intel 82550 Pro/100 Ethernet" },
{ 0x1229, 0x0f, "Intel 82551 Pro/100 Ethernet" },
{ 0x1229, 0x10, "Intel 82551 Pro/100 Ethernet" },
{ 0x1229, -1, "Intel 82557/8/9 Pro/100 Ethernet" },
{ 0x2449, -1, "Intel 82801BA/CAM (ICH2/3) Pro/100 Ethernet" },
{ 0, -1, 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);
static void fxp_intr(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_intr_body(struct fxp_softc *sc, struct ifnet *ifp,
u_int8_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);
All current uses of pci_set_powerstate are bogus, at least in theory. However, they are presently necessary due to bigger bogusness in the pci bus layer not doing the right thing on suspend/resume or on initial device probe. This is exactly the sort of thing that the BURN_BRIDGES option was invented for. Mark all of them as BURN_BRIDGES. As soon as I have the powerstate stuff properly integrated into the pci bus code, I intend to remove all these workarounds.
2003-07-03 14:00:57 +00:00
#ifndef BURN_BRIDGES
Reset the device's powerstate to d0 when resuming from a suspend operation, not just when we initally attach to the device. Submitted by: warner
2001-07-25 18:00:17 +00:00
static void fxp_powerstate_d0(device_t dev);
All current uses of pci_set_powerstate are bogus, at least in theory. However, they are presently necessary due to bigger bogusness in the pci bus layer not doing the right thing on suspend/resume or on initial device probe. This is exactly the sort of thing that the BURN_BRIDGES option was invented for. Mark all of them as BURN_BRIDGES. As soon as I have the powerstate stuff properly integrated into the pci bus code, I intend to remove all these workarounds.
2003-07-03 14:00:57 +00:00
#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 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);
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);
static void fxp_watchdog(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
static int fxp_add_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);
static u_int16_t fxp_eeprom_getword(struct fxp_softc *sc, int offset,
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,
u_int16_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);
static volatile int fxp_miibus_readreg(device_t dev, int phy, int reg);
static void fxp_miibus_writereg(device_t dev, int phy, int reg,
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);
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,
- 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
volatile u_int16_t *status, bus_dma_tag_t dmat,
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);
Sometimes cardbus attachments don't attach, so while we track down this problem put these lines back in. While they should be unnecessary, they appear to be sometimes necessary. Reviewed in concept: dfr Approved by: re (scottl@)
2003-11-28 05:28:29 +00:00
DRIVER_MODULE(fxp, cardbus, 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
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
{
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);
if (i == 0)
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),
CSR_READ_1(sc, FXP_CSR_SCB_RUSCUS),
CSR_READ_2(sc, FXP_CSR_FLOWCONTROL));
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
- 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(struct fxp_softc *sc, volatile u_int16_t *status,
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
{
int i = 10000;
- 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_sync(dmat, map, BUS_DMASYNC_POSTREAD);
while (!(le16toh(*status) & FXP_CB_STATUS_C) && --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);
- 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_sync(dmat, map, BUS_DMASYNC_POSTREAD);
}
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
}
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
/*
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
* Return identification string if this device is ours.
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
*/
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_probe(device_t dev)
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
u_int16_t devid;
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
u_int8_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)) {
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_set_desc(dev, ident->name);
return (0);
}
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
}
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
}
All current uses of pci_set_powerstate are bogus, at least in theory. However, they are presently necessary due to bigger bogusness in the pci bus layer not doing the right thing on suspend/resume or on initial device probe. This is exactly the sort of thing that the BURN_BRIDGES option was invented for. Mark all of them as BURN_BRIDGES. As soon as I have the powerstate stuff properly integrated into the pci bus code, I intend to remove all these workarounds.
2003-07-03 14:00:57 +00:00
#ifndef BURN_BRIDGES
Reset the device's powerstate to d0 when resuming from a suspend operation, not just when we initally attach to the device. Submitted by: warner
2001-07-25 18:00:17 +00:00
static void
fxp_powerstate_d0(device_t dev)
{
#if __FreeBSD_version >= 430002
u_int32_t iobase, membase, irq;
if (pci_get_powerstate(dev) != PCI_POWERSTATE_D0) {
/* Save important PCI config data. */
iobase = pci_read_config(dev, FXP_PCI_IOBA, 4);
membase = pci_read_config(dev, FXP_PCI_MMBA, 4);
irq = pci_read_config(dev, PCIR_INTLINE, 4);
/* Reset the power state. */
device_printf(dev, "chip is in D%d power mode "
"-- setting to D0\n", pci_get_powerstate(dev));
pci_set_powerstate(dev, PCI_POWERSTATE_D0);
/* Restore PCI config data. */
pci_write_config(dev, FXP_PCI_IOBA, iobase, 4);
pci_write_config(dev, FXP_PCI_MMBA, membase, 4);
pci_write_config(dev, PCIR_INTLINE, irq, 4);
}
#endif
}
All current uses of pci_set_powerstate are bogus, at least in theory. However, they are presently necessary due to bigger bogusness in the pci bus layer not doing the right thing on suspend/resume or on initial device probe. This is exactly the sort of thing that the BURN_BRIDGES option was invented for. Mark all of them as BURN_BRIDGES. As soon as I have the powerstate stuff properly integrated into the pci bus code, I intend to remove all these workarounds.
2003-07-03 14:00:57 +00:00
#endif
Reset the device's powerstate to d0 when resuming from a suspend operation, not just when we initally attach to the device. Submitted by: warner
2001-07-25 18:00:17 +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)
{
u_int32_t *addr;
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
{
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
int error = 0;
struct fxp_softc *sc = device_get_softc(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
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_rx *rxp;
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
u_int32_t val;
- 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
u_int16_t data, myea[ETHER_ADDR_LEN / 2];
Correct the maxsize and maxsegsz parameters for the bus_dma_tag_create() calls. Without this, fxp(4) + PAE would consume insane amounts of bounce pages.
2003-04-07 17:02:44 +00:00
int i, rid, m1, m2, prefer_iomap, maxtxseg;
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
int s, ipcbxmit_disable;
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
sc->dev = dev;
change timeout to be MPSAFE Sponsored by: FreeBSD Foundation
2003-09-05 22:37:31 +00:00
callout_init(&sc->stat_ch, CALLOUT_MPSAFE);
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);
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
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
s = splimp();
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
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);
All current uses of pci_set_powerstate are bogus, at least in theory. However, they are presently necessary due to bigger bogusness in the pci bus layer not doing the right thing on suspend/resume or on initial device probe. This is exactly the sort of thing that the BURN_BRIDGES option was invented for. Mark all of them as BURN_BRIDGES. As soon as I have the powerstate stuff properly integrated into the pci bus code, I intend to remove all these workarounds.
2003-07-03 14:00:57 +00:00
#ifndef BURN_BRIDGES
Reset the device's powerstate to d0 when resuming from a suspend operation, not just when we initally attach to the device. Submitted by: warner
2001-07-25 18:00:17 +00:00
fxp_powerstate_d0(dev);
All current uses of pci_set_powerstate are bogus, at least in theory. However, they are presently necessary due to bigger bogusness in the pci bus layer not doing the right thing on suspend/resume or on initial device probe. This is exactly the sort of thing that the BURN_BRIDGES option was invented for. Mark all of them as BURN_BRIDGES. As soon as I have the powerstate stuff properly integrated into the pci bus code, I intend to remove all these workarounds.
2003-07-03 14:00:57 +00:00
#endif
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
*/
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
m1 = PCIM_CMD_MEMEN;
m2 = PCIM_CMD_PORTEN;
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;
if (resource_int_value(device_get_name(dev), device_get_unit(dev),
"prefer_iomap", &prefer_iomap) == 0 && prefer_iomap != 0) {
m1 = PCIM_CMD_PORTEN;
m2 = PCIM_CMD_MEMEN;
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
}
- Don't call pci_enable_io() in drivers (unless needed for resume). - Don't test memory/port status and emit an error message; the PCI bus code will do this now.
2003-04-16 03:16:57 +00:00
sc->rtp = (m1 == PCIM_CMD_MEMEN)? SYS_RES_MEMORY : SYS_RES_IOPORT;
sc->rgd = (m1 == PCIM_CMD_MEMEN)? FXP_PCI_MMBA : FXP_PCI_IOBA;
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->mem = bus_alloc_resource_any(dev, sc->rtp, &sc->rgd, RF_ACTIVE);
- Don't call pci_enable_io() in drivers (unless needed for resume). - Don't test memory/port status and emit an error message; the PCI bus code will do this now.
2003-04-16 03:16:57 +00:00
if (sc->mem == NULL) {
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
sc->rtp =
(m2 == PCIM_CMD_MEMEN)? SYS_RES_MEMORY : SYS_RES_IOPORT;
sc->rgd = (m2 == PCIM_CMD_MEMEN)? FXP_PCI_MMBA : FXP_PCI_IOBA;
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->mem = bus_alloc_resource_any(dev, sc->rtp, &sc->rgd,
RF_ACTIVE);
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
}
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
if (!sc->mem) {
error = ENXIO;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
goto fail;
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
}
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
if (bootverbose) {
device_printf(dev, "using %s space register mapping\n",
sc->rtp == SYS_RES_MEMORY? "memory" : "I/O");
}
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
sc->sc_st = rman_get_bustag(sc->mem);
sc->sc_sh = rman_get_bushandle(sc->mem);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +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
/*
* Allocate our interrupt.
*/
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
rid = 0;
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
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
RF_SHAREABLE | RF_ACTIVE);
if (sc->irq == NULL) {
device_printf(dev, "could not map interrupt\n");
error = ENXIO;
goto fail;
}
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
/*
* Reset to a stable state.
*/
CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SELECTIVE_RESET);
DELAY(10);
/*
* Find out how large of an SEEPROM we have.
*/
fxp_autosize_eeprom(sc);
/*
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);
if ((data & FXP_PHY_DEVICE_MASK) != 0 &&
(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
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
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
Those sysctls shouldn't be writtable from inside a jail.
2004-05-26 23:03:27 +00:00
OID_AUTO, "int_delay", CTLTYPE_INT | CTLFLAG_RW,
'&' is not needed to pointerize functions. Spotted by: FlexeLint
2002-10-20 20:42:25 +00:00
&sc->tunable_int_delay, 0, sysctl_hw_fxp_int_delay, "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
"FXP driver receive interrupt microcode bundling delay");
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
SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
Those sysctls shouldn't be writtable from inside a jail.
2004-05-26 23:03:27 +00:00
OID_AUTO, "bundle_max", CTLTYPE_INT | CTLFLAG_RW,
'&' is not needed to pointerize functions. Spotted by: FlexeLint
2002-10-20 20:42:25 +00:00
&sc->tunable_bundle_max, 0, sysctl_hw_fxp_bundle_max, "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
"FXP driver receive interrupt microcode bundle size limit");
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
SYSCTL_ADD_INT(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "rnr", CTLFLAG_RD, &sc->rnr, 0,
"FXP RNR events");
SYSCTL_ADD_INT(device_get_sysctl_ctx(dev),
SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
OID_AUTO, "noflow", CTLFLAG_RW, &sc->tunable_noflow, 0,
"FXP flow control disabled");
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
/*
* Pull in device tunables.
*/
sc->tunable_int_delay = TUNABLE_INT_DELAY;
sc->tunable_bundle_max = TUNABLE_BUNDLE_MAX;
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->tunable_noflow = 0;
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
(void) resource_int_value(device_get_name(dev), device_get_unit(dev),
"int_delay", &sc->tunable_int_delay);
(void) resource_int_value(device_get_name(dev), device_get_unit(dev),
"bundle_max", &sc->tunable_bundle_max);
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
(void) resource_int_value(device_get_name(dev), device_get_unit(dev),
"noflow", &sc->tunable_noflow);
sc->rnr = 0;
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
/*
* Find out the chip revision; lump all 82557 revs together.
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
*/
fxp_read_eeprom(sc, &data, 5, 1);
if ((data >> 8) == 1)
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
sc->revision = FXP_REV_82557;
else
sc->revision = pci_get_revid(dev);
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
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
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
* a CU_RESUME command when it is entering the IDLE state. The
* 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
*/
i = pci_get_device(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
if (i == 0x2449 || (i > 0x1030 && i < 0x1039) ||
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 */
u_int16_t cksum;
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
}
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.
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
*
* At least some 82550 cards probed as "chip=0x12298086 rev=0x0d"
* truncate packets that end with an mbuf containing 1 to 3 bytes
* when used with this feature enabled in the previous version of the
* driver. This problem appears to be fixed now that the driver
* always sets the hardware parse bit in the IPCB structure, which
* the "Intel 8255x 10/100 Mbps Ethernet Controller Family Open
* Source Software Developer Manual" says is necessary in the
* cases where packet truncation was observed.
*
* The device hint "hint.fxp.UNIT_NUMBER.ipcbxmit_disable"
* allows this feature to be disabled at boot time.
*
* If fxp is not compiled into the kernel, this feature may also
* be disabled at run time:
* # kldunload fxp
* # kenv hint.fxp.0.ipcbxmit_disable=1
* # kldload fxp
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
*/
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
if (resource_int_value("fxp", device_get_unit(dev), "ipcbxmit_disable",
&ipcbxmit_disable) != 0)
ipcbxmit_disable = 0;
if (ipcbxmit_disable == 0 && (sc->revision == FXP_REV_82550 ||
sc->revision == FXP_REV_82550_C)) {
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;
} 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.
*/
Correct the maxsize and maxsegsz parameters for the bus_dma_tag_create() calls. Without this, fxp(4) + PAE would consume insane amounts of bounce pages.
2003-04-07 17:02:44 +00:00
maxtxseg = sc->flags & FXP_FLAG_EXT_RFA ? FXP_NTXSEG - 1 : FXP_NTXSEG;
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_dma_tag_create(NULL, 2, 0, BUS_SPACE_MAXADDR_32BIT,
Correct the maxsize and maxsegsz parameters for the bus_dma_tag_create() calls. Without this, fxp(4) + PAE would consume insane amounts of bounce pages.
2003-04-07 17:02:44 +00:00
BUS_SPACE_MAXADDR, NULL, NULL, MCLBYTES * maxtxseg,
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
maxtxseg, MCLBYTES, 0, busdma_lock_mutex, &Giant, &sc->fxp_mtag);
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, "could not allocate dma tag\n");
goto fail;
}
error = bus_dma_tag_create(NULL, 4, 0, BUS_SPACE_MAXADDR_32BIT,
BUS_SPACE_MAXADDR, NULL, NULL, sizeof(struct fxp_stats), 1,
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
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) {
device_printf(dev, "could not allocate dma tag\n");
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);
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)
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;
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) {
device_printf(dev, "could not map the stats buffer\n");
goto fail;
}
error = bus_dma_tag_create(NULL, 4, 0, BUS_SPACE_MAXADDR_32BIT,
BUS_SPACE_MAXADDR, NULL, NULL, FXP_TXCB_SZ, 1,
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
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) {
device_printf(dev, "could not allocate dma tag\n");
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);
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)
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;
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) {
device_printf(dev, "could not map DMA memory\n");
goto fail;
}
error = bus_dma_tag_create(NULL, 4, 0, BUS_SPACE_MAXADDR_32BIT,
BUS_SPACE_MAXADDR, NULL, NULL, sizeof(struct fxp_cb_mcs), 1,
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
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) {
device_printf(dev, "could not allocate dma tag\n");
goto fail;
}
error = bus_dmamem_alloc(sc->mcs_tag, (void **)&sc->mcsp,
BUS_DMA_NOWAIT, &sc->mcs_map);
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
goto fail;
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) {
device_printf(dev, "can't map the multicast setup command\n");
goto fail;
}
/*
* Pre-allocate the TX DMA maps.
*/
- 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++) {
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_create(sc->fxp_mtag, 0,
&sc->fxp_desc.tx_list[i].tx_map);
if (error) {
device_printf(dev, "can't create DMA map for TX\n");
goto fail;
}
}
error = bus_dmamap_create(sc->fxp_mtag, 0, &sc->spare_map);
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];
error = bus_dmamap_create(sc->fxp_mtag, 0, &rxp->rx_map);
if (error) {
device_printf(dev, "can't create DMA map for RX\n");
goto fail;
}
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 (fxp_add_rfabuf(sc, rxp) != 0) {
error = ENOMEM;
goto fail;
}
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);
sc->arpcom.ac_enaddr[0] = myea[0] & 0xff;
sc->arpcom.ac_enaddr[1] = myea[0] >> 8;
sc->arpcom.ac_enaddr[2] = myea[1] & 0xff;
sc->arpcom.ac_enaddr[3] = myea[1] >> 8;
sc->arpcom.ac_enaddr[4] = myea[2] & 0xff;
sc->arpcom.ac_enaddr[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
ifp = &sc->arpcom.ac_if;
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_baudrate = 100000000;
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;
ifp->if_watchdog = fxp_watchdog;
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;
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 checksum offload for 82550 or better chips */
if (sc->flags & FXP_FLAG_EXT_RFA) {
ifp->if_hwassist = FXP_CSUM_FEATURES;
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 |= IFCAP_HWCSUM;
ifp->if_capenable |= IFCAP_HWCSUM;
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
}
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;
ifp->if_capenable |= 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.
*/
ether_ifattach(ifp, sc->arpcom.ac_enaddr);
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 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
*/
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
ifp->if_snd.ifq_maxlen = FXP_NTXCB - 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
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.
* XXX This driver has been tested with the INTR_MPSAFFE flag set
* however, ifp and its functions are not fully locked so MPSAFE
* should not be used unless you can handle potential data loss.
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
*/
mark interrupt handlers MPSAFE
2003-10-29 18:32:14 +00:00
error = bus_setup_intr(dev, sc->irq, INTR_TYPE_NET | INTR_MPSAFE,
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_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");
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
ether_ifdetach(&sc->arpcom.ac_if);
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;
}
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:
splx(s);
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 (error)
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 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->irq)
bus_release_resource(sc->dev, SYS_RES_IRQ, 0, sc->irq);
if (sc->mem)
bus_release_resource(sc->dev, sc->rtp, sc->rgd, sc->mem);
- 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
if (sc->fxp_mtag) {
for (i = 0; i < FXP_NRFABUFS; i++) {
rxp = &sc->fxp_desc.rx_list[i];
if (rxp->rx_mbuf != NULL) {
bus_dmamap_sync(sc->fxp_mtag, rxp->rx_map,
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->fxp_mtag, rxp->rx_map);
m_freem(rxp->rx_mbuf);
}
bus_dmamap_destroy(sc->fxp_mtag, rxp->rx_map);
}
bus_dmamap_destroy(sc->fxp_mtag, 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
bus_dma_tag_destroy(sc->fxp_mtag);
- 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
}
if (sc->fxp_stag) {
for (i = 0; i < FXP_NTXCB; i++) {
txp = &sc->fxp_desc.tx_list[i];
if (txp->tx_mbuf != NULL) {
bus_dmamap_sync(sc->fxp_mtag, txp->tx_map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(sc->fxp_mtag, txp->tx_map);
m_freem(txp->tx_mbuf);
}
bus_dmamap_destroy(sc->fxp_mtag, 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_dma_tag_destroy(sc->fxp_stag);
- 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
}
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);
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);
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 s;
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 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);
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
s = splimp();
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
Fix tpyo in last commit. Noticed by: kuriyama-san
2003-04-30 04:14:56 +00:00
sc->suspended = 1; /* Do same thing as we do for suspend */
- 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
/*
* Close down routes etc.
*/
ether_ifdetach(&sc->arpcom.ac_if);
Changes to support NetBSD and the new ifmedia extensions. Submitted by: Jason Thorpe <thorpej@netbsd.org>
1997-09-05 10:23:58 +00:00
- 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);
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
- 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
*/
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
bus_teardown_intr(sc->dev, sc->irq, sc->ih);
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
splx(s);
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
/* 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
{
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.
*/
fxp_stop((struct fxp_softc *) 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
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
}
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);
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, s;
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);
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
s = splimp();
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);
Reset the device's powerstate to d0 when resuming from a suspend operation, not just when we initally attach to the device. Submitted by: warner
2001-07-25 18:00:17 +00:00
for (i = 0; i < 5; i++)
Use PCIR_BAR(x) instead of PCIR_MAPS. Glanced over by: imp, gibbs Tested by: i386 LINT
2003-09-02 17:30:40 +00:00
sc->saved_maps[i] = pci_read_config(dev, PCIR_BAR(i), 4);
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->saved_biosaddr = pci_read_config(dev, PCIR_BIOS, 4);
sc->saved_intline = pci_read_config(dev, PCIR_INTLINE, 1);
sc->saved_cachelnsz = pci_read_config(dev, PCIR_CACHELNSZ, 1);
sc->saved_lattimer = pci_read_config(dev, PCIR_LATTIMER, 1);
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
splx(s);
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
}
/*
* Device resume routine. Restore some PCI settings in case the BIOS
* doesn't, re-enable busmastering, and restart the interface if
* appropriate.
*/
static int
fxp_resume(device_t dev)
{
struct fxp_softc *sc = device_get_softc(dev);
struct ifnet *ifp = &sc->sc_if;
u_int16_t pci_command;
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, s;
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);
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
s = splimp();
All current uses of pci_set_powerstate are bogus, at least in theory. However, they are presently necessary due to bigger bogusness in the pci bus layer not doing the right thing on suspend/resume or on initial device probe. This is exactly the sort of thing that the BURN_BRIDGES option was invented for. Mark all of them as BURN_BRIDGES. As soon as I have the powerstate stuff properly integrated into the pci bus code, I intend to remove all these workarounds.
2003-07-03 14:00:57 +00:00
#ifndef BURN_BRIDGES
Reset the device's powerstate to d0 when resuming from a suspend operation, not just when we initally attach to the device. Submitted by: warner
2001-07-25 18:00:17 +00:00
fxp_powerstate_d0(dev);
All current uses of pci_set_powerstate are bogus, at least in theory. However, they are presently necessary due to bigger bogusness in the pci bus layer not doing the right thing on suspend/resume or on initial device probe. This is exactly the sort of thing that the BURN_BRIDGES option was invented for. Mark all of them as BURN_BRIDGES. As soon as I have the powerstate stuff properly integrated into the pci bus code, I intend to remove all these workarounds.
2003-07-03 14:00:57 +00:00
#endif
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
/* better way to do this? */
Reset the device's powerstate to d0 when resuming from a suspend operation, not just when we initally attach to the device. Submitted by: warner
2001-07-25 18:00:17 +00:00
for (i = 0; i < 5; i++)
Use PCIR_BAR(x) instead of PCIR_MAPS. Glanced over by: imp, gibbs Tested by: i386 LINT
2003-09-02 17:30:40 +00:00
pci_write_config(dev, PCIR_BAR(i), sc->saved_maps[i], 4);
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
pci_write_config(dev, PCIR_BIOS, sc->saved_biosaddr, 4);
pci_write_config(dev, PCIR_INTLINE, sc->saved_intline, 1);
pci_write_config(dev, PCIR_CACHELNSZ, sc->saved_cachelnsz, 1);
pci_write_config(dev, PCIR_LATTIMER, sc->saved_lattimer, 1);
/* reenable busmastering */
pci_command = pci_read_config(dev, PCIR_COMMAND, 2);
pci_command |= (PCIM_CMD_MEMEN|PCIM_CMD_BUSMASTEREN);
pci_write_config(dev, PCIR_COMMAND, pci_command, 2);
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
splx(s);
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
}
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_shiftin(struct fxp_softc *sc, int data, int length)
{
u_int16_t reg;
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
*/
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 u_int16_t
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
{
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
u_int16_t reg, data;
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
fxp_eeprom_putword(struct fxp_softc *sc, int offset, u_int16_t data)
{
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]);
}
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_txbuf(void *arg, bus_dma_segment_t *segs, int nseg,
bus_size_t mapsize, int error)
{
struct fxp_softc *sc;
struct fxp_cb_tx *txp;
int i;
if (error)
return;
KASSERT(nseg <= FXP_NTXSEG, ("too many DMA segments"));
sc = arg;
txp = sc->fxp_desc.tx_last->tx_next->tx_cb;
for (i = 0; i < nseg; i++) {
KASSERT(segs[i].ds_len <= MCLBYTES, ("segment size too large"));
/*
* 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) {
- 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
txp->tbd[i + 1].tb_addr = htole32(segs[i].ds_addr);
txp->tbd[i + 1].tb_size = htole32(segs[i].ds_len);
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
} else {
- 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
txp->tbd[i].tb_addr = htole32(segs[i].ds_addr);
txp->tbd[i].tb_size = htole32(segs[i].ds_len);
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->tbd_number = nseg;
}
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);
}
/*
* Start packet transmission on the interface.
* 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;
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
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 mbuf *mb_head;
int error;
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
/*
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
* See if we need to suspend xmit until the multicast filter
* has been reprogrammed (which can only be done at the head
* of the command chain).
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +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
if (sc->need_mcsetup) {
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
return;
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
}
Fixed a bug with the management of the pointer to the first TxCB in the ring that caused wrong things to happen sometimes. Doubled the number of transmit descriptors to 128 so that the internal FIFO in the NIC can be fully filled when dealing with small packets. Several minor performance improvements.
1996-09-22 11:48:54 +00:00
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
txp = NULL;
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
*/
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
while (ifp->if_snd.ifq_head != NULL && 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
*/
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
IF_DEQUEUE(&ifp->if_snd, mb_head);
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
/*
* Get pointer to next available tx desc.
*/
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_last->tx_next;
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
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
/*
* 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;
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
/*
* 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.
*/
if (mb_head->m_pkthdr.csum_flags) {
if (mb_head->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
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->tx_cb->ipcb_ip_schedule =
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
FXP_IPCB_TCPUDP_CHECKSUM_ENABLE;
if (mb_head->m_pkthdr.csum_flags & CSUM_TCP)
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->tx_cb->ipcb_ip_schedule |=
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
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 (mb_head->m_pkthdr.csum_flags & CSUM_IP) {
if (mb_head->m_pkthdr.len < 38) {
struct ip *ip;
mb_head->m_data += ETHER_HDR_LEN;
ip = mtod(mb_head, struct ip *);
ip->ip_sum = in_cksum(mb_head,
ip->ip_hl << 2);
mb_head->m_data -= ETHER_HDR_LEN;
} 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
txp->tx_cb->ipcb_ip_activation_high =
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
FXP_IPCB_HARDWAREPARSING_ENABLE;
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->tx_cb->ipcb_ip_schedule |=
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
FXP_IPCB_IP_CHECKSUM_ENABLE;
}
}
#endif
}
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
* Go through each of the mbufs in the chain and initialize
* the transmit buffer descriptors with the physical address
* and size of the mbuf.
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
error = bus_dmamap_load_mbuf(sc->fxp_mtag, txp->tx_map,
mb_head, fxp_dma_map_txbuf, sc, 0);
if (error && error != EFBIG) {
device_printf(sc->dev, "can't map mbuf (error %d)\n",
error);
m_freem(mb_head);
break;
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
}
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) {
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
struct mbuf *mn;
/*
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
* We ran out of segments. We have to recopy this
* mbuf chain first. Bail out if we can't get the
* new buffers.
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
*/
Replace if_fxp's handrolled defrag function with m_defrag. No actual bug with the existing code, just tidying things up. Submitted by: hmp
2003-06-02 05:16:42 +00:00
mn = m_defrag(mb_head, M_DONTWAIT);
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
if (mn == NULL) {
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
m_freem(mb_head);
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
break;
Replace if_fxp's handrolled defrag function with m_defrag. No actual bug with the existing code, just tidying things up. Submitted by: hmp
2003-06-02 05:16:42 +00:00
} else {
mb_head = mn;
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
}
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_mbuf(sc->fxp_mtag, txp->tx_map,
mb_head, fxp_dma_map_txbuf, sc, 0);
if (error) {
device_printf(sc->dev,
"can't map mbuf (error %d)\n", error);
m_freem(mb_head);
break;
}
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
}
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_sync(sc->fxp_mtag, txp->tx_map,
BUS_DMASYNC_PREWRITE);
txp->tx_mbuf = mb_head;
txp->tx_cb->cb_status = 0;
txp->tx_cb->byte_count = 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
if (sc->tx_queued != FXP_CXINT_THRESH - 1) {
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->tx_cb->cb_command =
- 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->tx_cmd | FXP_CB_COMMAND_SF |
FXP_CB_COMMAND_S);
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 {
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->tx_cb->cb_command =
- 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->tx_cmd | FXP_CB_COMMAND_SF |
FXP_CB_COMMAND_S | FXP_CB_COMMAND_I);
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
/*
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
* Set a 5 second timer just in case we don't hear
* from the card again.
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
*/
ifp->if_timer = 5;
}
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->tx_cb->tx_threshold = tx_threshold;
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +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__
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
/*
* 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.
Add an XXX comment for the atomic operation in the __alpha__ case so that it doesn't confuse more people. Use the more correct atomic_clear_16() instead of atomic_clear_short().
2003-04-08 18:35:51 +00:00
* XXX This is probably bogus and we're _not_ looking
* for atomicity here.
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
*/
Add an XXX comment for the atomic operation in the __alpha__ case so that it doesn't confuse more people. Use the more correct atomic_clear_16() instead of atomic_clear_short().
2003-04-08 18:35:51 +00:00
atomic_clear_16(&sc->fxp_desc.tx_last->tx_cb->cb_command,
- Endian fixes. - 6 -> ETHER_ADDR_LEN.
2003-04-08 17:21:15 +00:00
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
- Endian fixes. - 6 -> ETHER_ADDR_LEN.
2003-04-08 17:21:15 +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__*/
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_last = txp;
Fixed a bug with the management of the pointer to the first TxCB in the ring that caused wrong things to happen sometimes. Doubled the number of transmit descriptors to 128 so that the internal FIFO in the NIC can be fully filled when dealing with small packets. Several minor performance improvements.
1996-09-22 11:48:54 +00:00
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
/*
* Advance the beginning of the list forward if there are
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
* no other packets queued (when nothing is queued, tx_first
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
* sits on the last TxCB that was sent out).
*/
if (sc->tx_queued == 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
sc->fxp_desc.tx_first = 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
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
sc->tx_queued++;
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
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +00:00
/*
* Pass packet to bpf if there is a listener.
*/
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
BPF_MTAP(ifp, mb_head);
Rewrote fxp_start() for better clarity and efficiency.
1997-10-23 01:45:15 +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
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_PREWRITE);
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
* We're finished. If we added to the list, issue a RESUME to get DMA
* going again if suspended.
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
if (txp != NULL) {
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_CU_RESUME);
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
}
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;
static void
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;
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
u_int8_t 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
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);
Implemented per-interface polling(4) control.
2004-04-11 15:35:49 +00:00
if (!(ifp->if_capenable & IFCAP_POLLING)) {
ether_poll_deregister(ifp);
cmd = POLL_DEREGISTER;
}
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_DEREGISTER) { /* final call, enable interrupts */
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, 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);
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
return;
}
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) {
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
u_int8_t 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
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);
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
return; /* 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
}
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_intr_body(sc, ifp, statack, count);
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
}
#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;
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 ifnet *ifp = &sc->sc_if;
Fixed a bug with the management of the pointer to the first TxCB in the ring that caused wrong things to happen sometimes. Doubled the number of transmit descriptors to 128 so that the internal FIFO in the NIC can be fully filled when dealing with small packets. Several minor performance improvements.
1996-09-22 11:48:54 +00:00
u_int8_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
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 (ifp->if_flags & IFF_POLLING) {
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
}
Implemented per-interface polling(4) control.
2004-04-11 15:35:49 +00:00
if ((ifp->if_capenable & IFCAP_POLLING) &&
ether_poll_register(fxp_poll, ifp)) {
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
/* disable interrupts */
CSR_WRITE_1(sc, FXP_CSR_SCB_INTRCNTL, FXP_SCB_INTR_DISABLE);
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);
Unlock the fxp driver lock before calling fxp_poll() to avoid recursing on the lock when using DEVICE_POLLING. Tested by: Robin P. Blanchard <Robin.Blanchard@gactr.uga.edu> Reviewed by: njl
2003-07-11 20:49:36 +00:00
fxp_poll(ifp, 0, 1);
Fix (again) some blatent style bugs in DEVICE_POLLING code.
2001-12-15 02:41:52 +00:00
return;
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
* FXP_SCB_INTR_SWI bit always returns 0 on a read. If
* all bits are set, this may indicate that the card has
* been physically ejected, so ignore it.
*/
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);
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_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)
{
struct fxp_tx *txp;
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_PREREAD);
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) {
bus_dmamap_sync(sc->fxp_mtag, txp->tx_map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(sc->fxp_mtag, txp->tx_map);
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--;
}
sc->fxp_desc.tx_first = txp;
bus_dmamap_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_PREWRITE);
}
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
static void
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_intr_body(struct fxp_softc *sc, struct ifnet *ifp, u_int8_t statack,
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;
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);
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.
*/
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);
Revert rev 1.131. 1.131 is slightly broken, and I would commit the fix to that here, but it has been reported that any deviation from the original code is causing problems with some 82557 chips, causing them to lock hard. Until those issues have been figured out, going back to the original code is the best plan. Frustrated: Silby
2002-07-29 04:32:35 +00:00
ifp->if_timer = 0;
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
if (sc->tx_queued == 0) {
if (sc->need_mcsetup)
fxp_mc_setup(sc);
Revert rev 1.131. 1.131 is slightly broken, and I would commit the fix to that here, but it has been reported that any deviation from the original code is causing problems with some 82557 chips, causing them to lock hard. Until those issues have been figured out, going back to the original code is the best plan. Frustrated: Silby
2002-07-29 04:32:35 +00:00
}
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
* Try to start more packets transmitting.
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
if (ifp->if_snd.ifq_head != NULL)
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_start_body(ifp);
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
/*
* 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)
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
return;
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);
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_sync(sc->fxp_mtag, rxp->rx_map,
BUS_DMASYNC_POSTREAD);
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
- 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
if ((le16toh(rfa->rfa_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.
*/
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 (fxp_add_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;
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 -
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->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
le16toh(rfa->rfa_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
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
/* Do IP checksum checking. */
- 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
if (le16toh(rfa->rfa_status) & FXP_RFA_STATUS_PARSE) {
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 (rfa->rfax_csum_sts &
FXP_RFDX_CS_IP_CSUM_BIT_VALID)
m->m_pkthdr.csum_flags |=
CSUM_IP_CHECKED;
if (rfa->rfax_csum_sts &
FXP_RFDX_CS_IP_CSUM_VALID)
m->m_pkthdr.csum_flags |=
CSUM_IP_VALID;
if ((rfa->rfax_csum_sts &
FXP_RFDX_CS_TCPUDP_CSUM_BIT_VALID) &&
(rfa->rfax_csum_sts &
FXP_RFDX_CS_TCPUDP_CSUM_VALID)) {
m->m_pkthdr.csum_flags |=
CSUM_DATA_VALID|CSUM_PSEUDO_HDR;
m->m_pkthdr.csum_data = 0xffff;
}
}
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;
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);
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
}
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
}
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +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
/*
* 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;
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 ifnet *ifp = &sc->sc_if;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
struct fxp_stats *sp = sc->fxp_stats;
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 s;
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_LOCK(sc);
s = splimp();
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_sync(sc->fxp_stag, sc->fxp_smap, BUS_DMASYNC_POSTREAD);
- 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
ifp->if_opackets += le32toh(sp->tx_good);
ifp->if_collisions += le32toh(sp->tx_total_collisions);
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 (sp->rx_good) {
- 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
ifp->if_ipackets += le32toh(sp->rx_good);
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->rx_idle_secs = 0;
} else {
Defer rundown (m_freem) of completed transmit buffers for no longer than 1 second.
1998-10-22 02:00:49 +00:00
/*
* Receiver's been idle for another second.
*/
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->rx_idle_secs++;
}
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
ifp->if_ierrors +=
- 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
le32toh(sp->rx_crc_errors) +
le32toh(sp->rx_alignment_errors) +
le32toh(sp->rx_rnr_errors) +
le32toh(sp->rx_overrun_errors);
Implemented a better, dynamic, mechanism for adjusting the transmit threshold.
1996-09-19 09:15:20 +00:00
/*
* If any transmit underruns occured, bump up the transmit
* threshold by another 512 bytes (64 * 8).
*/
if (sp->tx_underruns) {
- 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
ifp->if_oerrors += le32toh(sp->tx_underruns);
Implemented a better, dynamic, mechanism for adjusting the transmit threshold.
1996-09-19 09:15:20 +00:00
if (tx_threshold < 192)
tx_threshold += 64;
}
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.
*/
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;
fxp_mc_setup(sc);
}
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
*/
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_sync(sc->fxp_stag, sc->fxp_smap,
BUS_DMASYNC_PREREAD);
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);
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
} else {
/*
* A previous command is still waiting to be accepted.
* Just zero our copy of the stats and wait for the
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
* next timer event to update them.
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
*/
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
sp->tx_good = 0;
Implemented a better, dynamic, mechanism for adjusting the transmit threshold.
1996-09-19 09:15:20 +00:00
sp->tx_underruns = 0;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
sp->tx_total_collisions = 0;
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
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
sp->rx_good = 0;
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
sp->rx_crc_errors = 0;
sp->rx_alignment_errors = 0;
sp->rx_rnr_errors = 0;
sp->rx_overrun_errors = 0;
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
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);
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);
splx(s);
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
{
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 ifnet *ifp = &sc->sc_if;
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
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
ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
ifp->if_timer = 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
#ifdef DEVICE_POLLING
ether_poll_deregister(ifp);
#endif
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 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
* Issue software reset, which also unloads the microcode.
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 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
sc->flags &= ~FXP_FLAG_UCODE;
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
CSR_WRITE_4(sc, FXP_CSR_PORT, FXP_PORT_SOFTWARE_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);
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
/*
* 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) {
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_sync(sc->fxp_mtag, txp[i].tx_map,
BUS_DMASYNC_POSTWRITE);
bus_dmamap_unload(sc->fxp_mtag, txp[i].tx_map);
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
}
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_sync(sc->cbl_tag, sc->cbl_map, 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
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_watchdog(struct ifnet *ifp)
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
struct fxp_softc *sc = ifp->if_softc;
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);
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");
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
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);
FXP_UNLOCK(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)
{
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 ifnet *ifp = &sc->sc_if;
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;
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 fxp_cb_mcs *mcsp;
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, prm, s;
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);
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
s = splimp();
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
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);
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_sync(sc->fxp_stag, sc->fxp_smap, BUS_DMASYNC_PREREAD);
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.
*/
if (ifp->if_flags & IFF_LINK0 && (sc->flags & FXP_FLAG_UCODE) == 0)
fxp_load_ucode(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
/*
* Initialize the multicast address list.
*/
if (fxp_mc_addrs(sc)) {
mcsp = sc->mcsp;
mcsp->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
mcsp->cb_command =
htole16(FXP_CB_COMMAND_MCAS | FXP_CB_COMMAND_EL);
mcsp->link_addr = 0xffffffff;
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
/*
* Start the multicast setup command.
*/
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
bus_dmamap_sync(sc->mcs_tag, sc->mcs_map, BUS_DMASYNC_PREWRITE);
CSR_WRITE_4(sc, FXP_CSR_SCB_GENERAL, sc->mcs_addr);
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
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, &mcsp->cb_status, sc->mcs_tag, sc->mcs_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_dmamap_sync(sc->mcs_tag, sc->mcs_map,
BUS_DMASYNC_POSTWRITE);
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 */
cbp->dyn_tbd = 0; /* (no) dynamic TBD mode */
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 */
cbp->tcp_udp_cksum = 0; /* (don't) enable checksum */
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 */
cbp->magic_pkt_dis = 0; /* (don't) disable magic packet */
/* must set wake_en in PMCSR also */
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 */
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->mc_all = sc->flags & FXP_FLAG_ALL_MCAST ? 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->gamla_rx = sc->flags & FXP_FLAG_EXT_RFA ? 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) */
}
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);
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_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_PREWRITE);
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);
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_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_POSTWRITE);
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;
Remove all the bogus volatile qualifiers from the structs definitions and associated evil casts to discard them.
2003-04-03 14:08:35 +00:00
bcopy(sc->arpcom.ac_enaddr, cb_ias->macaddr,
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
sizeof(sc->arpcom.ac_enaddr));
/*
* 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);
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_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_PREWRITE);
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);
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_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_POSTWRITE);
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_cb = tcbp + i;
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);
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_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_PREWRITE);
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);
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
ifp->if_flags |= IFF_RUNNING;
ifp->if_flags &= ~IFF_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!
*/
Increase size of ifnet.if_flags from 16 bits (short) to 32 bits (int). To avoid breaking application ABI use unused ifreq.ifru_flags[1] for upper 16 bits in SIOCSIFFLAGS and SIOCGIFFLAGS ioctl's. Reviewed by: -hackers, -net
2002-08-18 07:05:00 +00:00
if ( ifp->if_flags & IFF_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);
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
splx(s);
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);
mii_mediachg(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
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);
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
if (ifmr->ifm_status & IFM_10_T && sc->flags & FXP_FLAG_CU_RESUME_BUG)
sc->cu_resume_bug = 1;
else
sc->cu_resume_bug = 0;
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
* adding the 'oldm' (if non-NULL) on to the end of the list -
Seventy-odd "its" / "it's" typos in comments fixed as per kern/6108.
1998-04-17 22:37:19 +00:00
* tossing out its old contents and recycling it.
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
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_add_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;
struct fxp_rfa *rfa, *p_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
struct fxp_rx *p_rx;
bus_dmamap_t tmp_map;
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);
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 (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;
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. */
error = bus_dmamap_load(sc->fxp_mtag, sc->spare_map, rfa,
MCLBYTES - RFA_ALIGNMENT_FUDGE, fxp_dma_map_addr,
&rxp->rx_addr, 0);
if (error) {
m_freem(m);
return (error);
}
bus_dmamap_unload(sc->fxp_mtag, rxp->rx_map);
tmp_map = sc->spare_map;
sc->spare_map = rxp->rx_map;
rxp->rx_map = tmp_map;
rxp->rx_mbuf = m;
- 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_dmamap_sync(sc->fxp_mtag, rxp->rx_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
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;
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_sync(sc->fxp_mtag, p_rx->rx_map,
- 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
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;
return (0);
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +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
static volatile 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
}
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_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");
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;
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
int flag, mask, s, error = 0;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +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
/*
* Detaching causes us to call ioctl with the mutex owned. Preclude
* that by saying we're busy if the lock is already held.
*/
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
if (FXP_LOCKED(sc))
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
return (EBUSY);
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
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);
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
s = splimp();
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
switch (command) {
case SIOCSIFFLAGS:
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 (ifp->if_flags & IFF_ALLMULTI)
sc->flags |= FXP_FLAG_ALL_MCAST;
else
sc->flags &= ~FXP_FLAG_ALL_MCAST;
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) {
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
} else {
if (ifp->if_flags & IFF_RUNNING)
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
}
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
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 (ifp->if_flags & IFF_ALLMULTI)
sc->flags |= FXP_FLAG_ALL_MCAST;
else
sc->flags &= ~FXP_FLAG_ALL_MCAST;
Device driver for Intel Pro/100 PCI Fast Ethernet controller.
1995-11-28 23:55:26 +00:00
/*
Use the new if_multiaddrs list for multicast addresses rather than the previous hackery involving struct in_ifaddr and arpcom. Get rid of the abominable multi_kludge. Update all network interfaces to use the new machanism. Distressingly few Ethernet drivers program the multicast filter properly (assuming the hardware has one, which it usually does).
1997-01-13 21:26:53 +00:00
* Multicast list has changed; set the hardware filter
* accordingly.
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->flags & FXP_FLAG_ALL_MCAST) == 0)
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
fxp_mc_setup(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
* fxp_mc_setup() can set FXP_FLAG_ALL_MCAST, so check it
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
* again rather than else {}.
*/
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->flags & FXP_FLAG_ALL_MCAST)
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);
Use the new if_multiaddrs list for multicast addresses rather than the previous hackery involving struct in_ifaddr and arpcom. Get rid of the abominable multi_kludge. Update all network interfaces to use the new machanism. Distressingly few Ethernet drivers program the multicast filter properly (assuming the hardware has one, which it usually does).
1997-01-13 21:26:53 +00:00
error = 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
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:
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;
if (mask & IFCAP_POLLING)
ifp->if_capenable ^= IFCAP_POLLING;
if (mask & IFCAP_VLAN_MTU) {
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)
fxp_init_body(sc);
}
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:
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
/*
* ether_ioctl() will eventually call fxp_start() which
* will result in mutex recursion so drop it first.
*/
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
error = ether_ioctl(ifp, command, data);
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
if (FXP_LOCKED(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
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
splx(s);
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;
struct ifnet *ifp = &sc->sc_if;
struct ifmultiaddr *ifma;
int nmcasts;
nmcasts = 0;
if ((sc->flags & FXP_FLAG_ALL_MCAST) == 0) {
#if __FreeBSD_version < 500000
LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
#else
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
#endif
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
if (nmcasts >= MAXMCADDR) {
sc->flags |= FXP_FLAG_ALL_MCAST;
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++;
}
}
- 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.
*
* This function must be called at splimp.
*/
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
{
struct fxp_cb_mcs *mcsp = sc->mcsp;
struct ifnet *ifp = &sc->sc_if;
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;
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);
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
/*
* If there are queued commands, we must wait until they are all
* completed. If we are already waiting, then add a NOP command
* with interrupt option so that we're notified when all commands
* have been completed - fxp_start() ensures that no additional
* TX commands will be added when need_mcsetup is true.
*/
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 (sc->tx_queued) {
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
/*
* need_mcsetup will be true if we are already waiting for the
* NOP command to be completed (see below). In this case, bail.
*/
if (sc->need_mcsetup)
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
sc->need_mcsetup = 1;
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
/*
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 a NOP command with interrupt so that we are notified
* when all TX commands have been processed.
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(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_last->tx_next;
txp->tx_mbuf = NULL;
txp->tx_cb->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
txp->tx_cb->cb_command = htole16(FXP_CB_COMMAND_NOP |
FXP_CB_COMMAND_S | FXP_CB_COMMAND_I);
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
/*
* Advance the end of list forward.
*/
- 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
sc->fxp_desc.tx_last->tx_cb->cb_command &=
htole16(~FXP_CB_COMMAND_S);
Move a bus_dmamap_sync() to the correct place.
2003-04-06 23:16:00 +00:00
bus_dmamap_sync(sc->cbl_tag, sc->cbl_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
sc->fxp_desc.tx_last = txp;
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
sc->tx_queued++;
/*
* Issue a resume in case the CU has just suspended.
*/
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_CU_RESUME);
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
/*
* Set a 5 second timer just in case we don't hear from the
* card again.
*/
ifp->if_timer = 5;
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
return;
}
sc->need_mcsetup = 0;
/*
* Initialize multicast setup descriptor.
*/
mcsp->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
mcsp->cb_command = htole16(FXP_CB_COMMAND_MCAS |
FXP_CB_COMMAND_S | FXP_CB_COMMAND_I);
mcsp->link_addr = htole32(sc->fxp_desc.cbl_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
txp = &sc->fxp_desc.mcs_tx;
txp->tx_mbuf = NULL;
txp->tx_cb = (struct fxp_cb_tx *)sc->mcsp;
txp->tx_next = sc->fxp_desc.tx_list;
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
(void) fxp_mc_addrs(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
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;
/*
* Wait until command unit is not active. This should never
* be the case when nothing is queued, but make sure anyway.
*/
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;
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
while ((CSR_READ_1(sc, FXP_CSR_SCB_RUSCUS) >> 6) ==
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_SCB_CUS_ACTIVE && --count)
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);
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_sync(sc->mcs_tag, sc->mcs_map, BUS_DMASYNC_PREWRITE);
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);
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
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
ifp->if_timer = 2;
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
return;
}
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 u_int32_t fxp_ucode_d101a[] = D101_A_RCVBUNDLE_UCODE;
static u_int32_t fxp_ucode_d101b0[] = D101_B0_RCVBUNDLE_UCODE;
static u_int32_t fxp_ucode_d101ma[] = D101M_B_RCVBUNDLE_UCODE;
static u_int32_t fxp_ucode_d101s[] = D101S_RCVBUNDLE_UCODE;
static u_int32_t fxp_ucode_d102[] = D102_B_RCVBUNDLE_UCODE;
static u_int32_t fxp_ucode_d102c[] = D102_C_RCVBUNDLE_UCODE;
#define UCODE(x) x, sizeof(x)
struct ucode {
u_int32_t revision;
u_int32_t *ucode;
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 },
{ 0, NULL, 0, 0, 0 }
};
static void
fxp_load_ucode(struct fxp_softc *sc)
{
struct ucode *uc;
struct fxp_cb_ucode *cbp;
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 */
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
memcpy(cbp->ucode, uc->ucode, uc->length);
if (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
*(u_int16_t *)&cbp->ucode[uc->int_delay_offset] =
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)
- 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
*(u_int16_t *)&cbp->ucode[uc->bundle_max_offset] =
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);
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_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_PREWRITE);
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);
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_sync(sc->cbl_tag, sc->cbl_map, BUS_DMASYNC_POSTWRITE);
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",
sc->tunable_int_delay,
uc->bundle_max_offset == 0 ? 0 : sc->tunable_bundle_max);
sc->flags |= FXP_FLAG_UCODE;
}
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
* to convert this to the appropriate clock ticks before using.
*/
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