d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
f1650f412f
freebsd-skq/sys/dev/bge/if_bge.c

3848 lines
102 KiB
C
Raw Normal View History

Start each of the license/copyright comments with /*-, minor shuffle of lines
2005-01-06 01:43:34 +00:00
/*-
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* Copyright (c) 2001 Wind River Systems
* Copyright (c) 1997, 1998, 1999, 2001
* Bill Paul <wpaul@windriver.com>. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Bill Paul.
* 4. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY Bill Paul 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 Bill Paul OR THE VOICES IN HIS HEAD
* 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("$FreeBSD$");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Broadcom BCM570x family gigabit ethernet driver for FreeBSD.
Use __FBSDID rather than rcsid[].
2003-04-03 21:36:33 +00:00
*
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* The Broadcom BCM5700 is based on technology originally developed by
* Alteon Networks as part of the Tigon I and Tigon II gigabit ethernet
* MAC chips. The BCM5700, sometimes refered to as the Tigon III, has
* two on-board MIPS R4000 CPUs and can have as much as 16MB of external
* SSRAM. The BCM5700 supports TCP, UDP and IP checksum offload, jumbo
* frames, highly configurable RX filtering, and 16 RX and TX queues
* (which, along with RX filter rules, can be used for QOS applications).
* Other features, such as TCP segmentation, may be available as part
* of value-added firmware updates. Unlike the Tigon I and Tigon II,
* firmware images can be stored in hardware and need not be compiled
* into the driver.
*
* The BCM5700 supports the PCI v2.2 and PCI-X v1.0 standards, and will
* function in a 32-bit/64-bit 33/66Mhz bus, or a 64-bit/133Mhz bus.
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
*
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* The BCM5701 is a single-chip solution incorporating both the BCM5700
Teach the bge driver about the BCM5701 (specifically, the 3Com 3c996B-T, with the 5701 rev B5 ASIC). One thing that confuses me still is that the 'link state change' bit in the status block seems to change state an awful lot. I have a workaround for this in place now, but it needs more investigation. For the moment though, this is enough to get the driver to work with this card.
2002-03-22 06:45:40 +00:00
* MAC and a BCM5401 10/100/1000 PHY. Unlike the BCM5700, the BCM5701
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* does not support external SSRAM.
*
* Broadcom also produces a variation of the BCM5700 under the "Altima"
* brand name, which is functionally similar but lacks PCI-X support.
*
* Without external SSRAM, you can only have at most 4 TX rings,
* and the use of the mini RX ring is disabled. This seems to imply
* that these features are simply not available on the BCM5701. As a
* result, this driver does not implement any support for the mini RX
* ring.
*/
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef HAVE_KERNEL_OPTION_HEADERS
#include "opt_device_polling.h"
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include <sys/param.h>
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
#include <sys/endian.h>
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include <sys/systm.h>
#include <sys/sockio.h>
#include <sys/mbuf.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
Add missing <sys/module.h> includes
2004-05-30 20:08:47 +00:00
#include <sys/module.h>
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include <sys/socket.h>
#include <net/if.h>
#include <net/if_arp.h>
#include <net/ethernet.h>
#include <net/if_dl.h>
#include <net/if_media.h>
#include <net/bpf.h>
#include <net/if_types.h>
#include <net/if_vlan_var.h>
#include <netinet/in_systm.h>
#include <netinet/in.h>
#include <netinet/ip.h>
#include <machine/clock.h> /* for DELAY */
#include <machine/bus.h>
#include <machine/resource.h>
#include <sys/bus.h>
#include <sys/rman.h>
#include <dev/mii/mii.h>
#include <dev/mii/miivar.h>
Remove miidevs.h and generate it from miidevs at compile time. The devlist2h.awk tool to do this has been repocopied to sys/tools/.
2003-01-19 02:59:34 +00:00
#include "miidevs.h"
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include <dev/mii/brgphyreg.h>
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/pcireg.h>
#include <dev/pci/pcivar.h>
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include <dev/bge/if_bgereg.h>
Backout the change I made before 5.4-R since I wasn't aware that it was only a problem with one particular switch module. Create a kernel option BGE_FAKE_AUTONEG that restores the 5.4 behavior, which should make the DNLK switch module work. IBM/Intel blades with Intel or AD switch modules should work without patching or kernel options with this commit. Hardware for testing provided by several folks, including Danny Braniss <danny@cs.huji.ac.il>, Achim Patzner <ap@bnc.net>, and OffMyServer. Approved by: re
2005-06-24 21:43:47 +00:00
#include "opt_bge.h"
Disable hardware checksum offloading of IP fragments in the bge driver. I tried a few obvious experiments, but was unable to make the 3c996B-T generate correct UDP checksums for transmitted fragmented packets. I'm not so sure the device is even capable of it. This fixes NFS over UDP. MFC after: 1 day
2002-07-04 23:19:20 +00:00
#define BGE_CSUM_FEATURES (CSUM_IP | CSUM_TCP | CSUM_UDP)
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
#define ETHER_MIN_NOPAD (ETHER_MIN_LEN - ETHER_CRC_LEN) /* i.e., 60 */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +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(bge, pci, 1, 1, 1);
MODULE_DEPEND(bge, ether, 1, 1, 1);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
MODULE_DEPEND(bge, miibus, 1, 1, 1);
Replace FreeBSD 3.x syntax (controller miibus0) with 4.x syntax (device miibus) in time for 7.0 :-)
2005-10-22 05:06:55 +00:00
/* "device miibus" required. See GENERIC if you get errors here. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include "miibus_if.h"
/*
* Various supported device vendors/types and their names. Note: the
* spec seems to indicate that the hardware still has Alteon's vendor
* ID burned into it, though it will always be overriden by the vendor
* ID in the EEPROM. Just to be safe, we cover all possibilities.
*/
Put the ASIC revision into the device description string so it appears in the dmesg output. MFC after: 3 days
2003-01-07 01:33:03 +00:00
#define BGE_DEVDESC_MAX 64 /* Maximum device description length */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static struct bge_type bge_devs[] = {
{ ALT_VENDORID, ALT_DEVICEID_BCM5700,
"Broadcom BCM5700 Gigabit Ethernet" },
{ ALT_VENDORID, ALT_DEVICEID_BCM5701,
"Broadcom BCM5701 Gigabit Ethernet" },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5700,
"Broadcom BCM5700 Gigabit Ethernet" },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5701,
"Broadcom BCM5701 Gigabit Ethernet" },
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5702,
"Broadcom BCM5702 Gigabit Ethernet" },
Add support for the BCM5702x chips. Reviewed by: jdp MFC after: 1 week
2002-09-28 18:04:12 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5702X,
"Broadcom BCM5702X Gigabit Ethernet" },
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5703,
"Broadcom BCM5703 Gigabit Ethernet" },
Add support for the BCM5703x chips. I do not have one of these cards to test; however the submitter reports that this patch works with the on-board interface on the IBM x235 server. Submitted by: Jung-uk Kim <jkim@niksun.com> MFC after: 1 month
2002-09-08 19:12:02 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5703X,
"Broadcom BCM5703X Gigabit Ethernet" },
Add untested support for the Broadcom 5704. Obtained from: NetBSD
2003-05-02 19:53:40 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5704C,
"Broadcom BCM5704C Dual Gigabit Ethernet" },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5704S,
"Broadcom BCM5704S Dual Gigabit Ethernet" },
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705,
"Broadcom BCM5705 Gigabit Ethernet" },
Add support for BCM5705K Submitted by: candy@kgc.co.jp PR: kern/67110
2004-05-24 04:46:54 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705K,
"Broadcom BCM5705K Gigabit Ethernet" },
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705M,
"Broadcom BCM5705M Gigabit Ethernet" },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705M_ALT,
"Broadcom BCM5705M Gigabit Ethernet" },
Support the 5714C Submitted by: John Cagle <john dot cagle at hp dot com>
2005-05-19 21:08:59 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5714C,
"Broadcom BCM5714C Gigabit Ethernet" },
Add support for the 5721 which seems to be similar to the 5750/5751. Tested by: Vivek Khera vivek at khera dot org
2004-12-17 21:50:18 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5721,
"Broadcom BCM5721 Gigabit Ethernet" },
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5750,
"Broadcom BCM5750 Gigabit Ethernet" },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5750M,
"Broadcom BCM5750M Gigabit Ethernet" },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5751,
"Broadcom BCM5751 Gigabit Ethernet" },
Adding new device ID for BCM5751M support. Submitted by: FUJITA Kazutoshi <fujita at soum dot co dot jp> Reviewed by: sam (mentor)
2005-03-12 06:51:25 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5751M,
"Broadcom BCM5751M Gigabit Ethernet" },
Recognize Broadcom BCM5752 chip, that can be found in HP DC7600. PR: kern/88940 Submitted by: Alexander Hausner
2005-11-15 14:43:22 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5752,
"Broadcom BCM5752 Gigabit Ethernet" },
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5782,
"Broadcom BCM5782 Gigabit Ethernet" },
Add support for the Broadcom BCM5788 (just another entry to the PCI device ID list, probably a 5705 ASIC). Submitted by: Marcel Prisi <marcel@virtua.ch>
2003-10-31 16:25:39 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5788,
"Broadcom BCM5788 Gigabit Ethernet" },
Add PCI ID for BCM5789. Submitted by: S. Aeschbacher
2005-05-22 03:16:45 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5789,
"Broadcom BCM5789 Gigabit Ethernet" },
Add support for the Broadcom BCM5901 and BCM5901 rev A2 chips. These are 10/100 only NICs found on the IBM Thinkpad R40E and G40. These seem to be based on the BCM5705 MAC but with a PHY that doesn't support 1000Mbps modes. Submitted by: Igor Sviridov <sia@nest.org>
2003-08-12 05:18:51 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5901,
"Broadcom BCM5901 Fast Ethernet" },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5901A2,
"Broadcom BCM5901A2 Fast Ethernet" },
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{ SK_VENDORID, SK_DEVICEID_ALTIMA,
"SysKonnect Gigabit Ethernet" },
Add support for the Netgear GA302T 10/100/1000 adapter. Given that it's a 32-bit card, it's quite nice for $75. MFC after: 3 days
2002-01-27 01:00:16 +00:00
{ ALTIMA_VENDORID, ALTIMA_DEVICE_AC1000,
"Altima AC1000 Gigabit Ethernet" },
Add the PCI ID for yet another bge chip: the Altima 1002. Submitted by: Pavel Gubin <pg@rainbow.ie.tusur.ru>
2004-01-08 17:19:11 +00:00
{ ALTIMA_VENDORID, ALTIMA_DEVICE_AC1002,
"Altima AC1002 Gigabit Ethernet" },
Recognize the Altima AC9100 chip, which is used in current versions of the Netgear GA302-T. I changed the symbolic names in the submitter's patch to reflect the part number of the chip instead of the board. PR: kern/38988 Submitted by: Brad Chapman <chapmanb@arches.uga.edu> MFC after: 2 days
2002-07-26 03:47:08 +00:00
{ ALTIMA_VENDORID, ALTIMA_DEVICE_AC9100,
"Altima AC9100 Gigabit Ethernet" },
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{ 0, 0, NULL }
};
Remove __P.
2002-03-20 02:08:01 +00:00
static int bge_probe (device_t);
static int bge_attach (device_t);
static int bge_detach (device_t);
Implement suspend/resume methods to be more ACPI friendly. I'm able to suspend/resume my laptop without this change, but then I need to wait for the watchdog to reset the card. With this change, it is ready immediately. Glanced at by: glebius
2005-09-28 19:20:49 +00:00
static int bge_suspend (device_t);
static int bge_resume (device_t);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static void bge_release_resources
Remove __P.
2002-03-20 02:08:01 +00:00
(struct bge_softc *);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
static void bge_dma_map_addr (void *, bus_dma_segment_t *, int, int);
static int bge_dma_alloc (device_t);
static void bge_dma_free (struct bge_softc *);
Remove __P.
2002-03-20 02:08:01 +00:00
static void bge_txeof (struct bge_softc *);
static void bge_rxeof (struct bge_softc *);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
static void bge_tick_locked (struct bge_softc *);
Remove __P.
2002-03-20 02:08:01 +00:00
static void bge_tick (void *);
static void bge_stats_update (struct bge_softc *);
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
static void bge_stats_update_regs
(struct bge_softc *);
Remove __P.
2002-03-20 02:08:01 +00:00
static int bge_encap (struct bge_softc *, struct mbuf *,
u_int32_t *);
static void bge_intr (void *);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
static void bge_start_locked (struct ifnet *);
Remove __P.
2002-03-20 02:08:01 +00:00
static void bge_start (struct ifnet *);
static int bge_ioctl (struct ifnet *, u_long, caddr_t);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
static void bge_init_locked (struct bge_softc *);
Remove __P.
2002-03-20 02:08:01 +00:00
static void bge_init (void *);
static void bge_stop (struct bge_softc *);
static void bge_watchdog (struct ifnet *);
static void bge_shutdown (device_t);
static int bge_ifmedia_upd (struct ifnet *);
static void bge_ifmedia_sts (struct ifnet *, struct ifmediareq *);
static u_int8_t bge_eeprom_getbyte (struct bge_softc *, int, u_int8_t *);
static int bge_read_eeprom (struct bge_softc *, caddr_t, int, int);
static void bge_setmulti (struct bge_softc *);
static int bge_newbuf_std (struct bge_softc *, int, struct mbuf *);
static int bge_newbuf_jumbo (struct bge_softc *, int, struct mbuf *);
static int bge_init_rx_ring_std (struct bge_softc *);
static void bge_free_rx_ring_std (struct bge_softc *);
static int bge_init_rx_ring_jumbo (struct bge_softc *);
static void bge_free_rx_ring_jumbo (struct bge_softc *);
static void bge_free_tx_ring (struct bge_softc *);
static int bge_init_tx_ring (struct bge_softc *);
static int bge_chipinit (struct bge_softc *);
static int bge_blockinit (struct bge_softc *);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Fix warnings: the bge_bpd_read*() functions are not used (#ifdef notdef)
2001-12-18 08:03:25 +00:00
#ifdef notdef
Remove __P.
2002-03-20 02:08:01 +00:00
static u_int8_t bge_vpd_readbyte(struct bge_softc *, int);
static void bge_vpd_read_res (struct bge_softc *, struct vpd_res *, int);
static void bge_vpd_read (struct bge_softc *);
Fix warnings: the bge_bpd_read*() functions are not used (#ifdef notdef)
2001-12-18 08:03:25 +00:00
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static u_int32_t bge_readmem_ind
Remove __P.
2002-03-20 02:08:01 +00:00
(struct bge_softc *, int);
static void bge_writemem_ind (struct bge_softc *, int, int);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#ifdef notdef
static u_int32_t bge_readreg_ind
Remove __P.
2002-03-20 02:08:01 +00:00
(struct bge_softc *, int);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#endif
Remove __P.
2002-03-20 02:08:01 +00:00
static void bge_writereg_ind (struct bge_softc *, int, int);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Remove __P.
2002-03-20 02:08:01 +00:00
static int bge_miibus_readreg (device_t, int, int);
static int bge_miibus_writereg (device_t, int, int, int);
static void bge_miibus_statchg (device_t);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
static void bge_poll (struct ifnet *ifp, enum poll_cmd cmd,
int count);
static void bge_poll_locked (struct ifnet *ifp, enum poll_cmd cmd,
int count);
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Remove __P.
2002-03-20 02:08:01 +00:00
static void bge_reset (struct bge_softc *);
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
static void bge_link_upd (struct bge_softc *);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static device_method_t bge_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, bge_probe),
DEVMETHOD(device_attach, bge_attach),
DEVMETHOD(device_detach, bge_detach),
DEVMETHOD(device_shutdown, bge_shutdown),
Implement suspend/resume methods to be more ACPI friendly. I'm able to suspend/resume my laptop without this change, but then I need to wait for the watchdog to reset the card. With this change, it is ready immediately. Glanced at by: glebius
2005-09-28 19:20:49 +00:00
DEVMETHOD(device_suspend, bge_suspend),
DEVMETHOD(device_resume, bge_resume),
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* bus interface */
DEVMETHOD(bus_print_child, bus_generic_print_child),
DEVMETHOD(bus_driver_added, bus_generic_driver_added),
/* MII interface */
DEVMETHOD(miibus_readreg, bge_miibus_readreg),
DEVMETHOD(miibus_writereg, bge_miibus_writereg),
DEVMETHOD(miibus_statchg, bge_miibus_statchg),
{ 0, 0 }
};
static driver_t bge_driver = {
"bge",
bge_methods,
sizeof(struct bge_softc)
};
static devclass_t bge_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(bge, pci, bge_driver, bge_devclass, 0, 0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
DRIVER_MODULE(miibus, bge, miibus_driver, miibus_devclass, 0, 0);
static u_int32_t
bge_readmem_ind(sc, off)
struct bge_softc *sc;
int off;
{
device_t dev;
dev = sc->bge_dev;
pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, off, 4);
return(pci_read_config(dev, BGE_PCI_MEMWIN_DATA, 4));
}
static void
bge_writemem_ind(sc, off, val)
struct bge_softc *sc;
int off, val;
{
device_t dev;
dev = sc->bge_dev;
pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, off, 4);
pci_write_config(dev, BGE_PCI_MEMWIN_DATA, val, 4);
return;
}
#ifdef notdef
static u_int32_t
bge_readreg_ind(sc, off)
struct bge_softc *sc;
int off;
{
device_t dev;
dev = sc->bge_dev;
pci_write_config(dev, BGE_PCI_REG_BASEADDR, off, 4);
return(pci_read_config(dev, BGE_PCI_REG_DATA, 4));
}
#endif
static void
bge_writereg_ind(sc, off, val)
struct bge_softc *sc;
int off, val;
{
device_t dev;
dev = sc->bge_dev;
pci_write_config(dev, BGE_PCI_REG_BASEADDR, off, 4);
pci_write_config(dev, BGE_PCI_REG_DATA, val, 4);
return;
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/*
* Map a single buffer address.
*/
static void
bge_dma_map_addr(arg, segs, nseg, error)
void *arg;
bus_dma_segment_t *segs;
int nseg;
int error;
{
struct bge_dmamap_arg *ctx;
if (error)
return;
ctx = arg;
if (nseg > ctx->bge_maxsegs) {
ctx->bge_maxsegs = 0;
return;
}
ctx->bge_busaddr = segs->ds_addr;
return;
}
Fix warnings: the bge_bpd_read*() functions are not used (#ifdef notdef)
2001-12-18 08:03:25 +00:00
#ifdef notdef
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static u_int8_t
bge_vpd_readbyte(sc, addr)
struct bge_softc *sc;
int addr;
{
int i;
device_t dev;
u_int32_t val;
dev = sc->bge_dev;
pci_write_config(dev, BGE_PCI_VPD_ADDR, addr, 2);
for (i = 0; i < BGE_TIMEOUT * 10; i++) {
DELAY(10);
if (pci_read_config(dev, BGE_PCI_VPD_ADDR, 2) & BGE_VPD_FLAG)
break;
}
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "VPD read timed out\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(0);
}
val = pci_read_config(dev, BGE_PCI_VPD_DATA, 4);
return((val >> ((addr % 4) * 8)) & 0xFF);
}
static void
bge_vpd_read_res(sc, res, addr)
struct bge_softc *sc;
struct vpd_res *res;
int addr;
{
int i;
u_int8_t *ptr;
ptr = (u_int8_t *)res;
for (i = 0; i < sizeof(struct vpd_res); i++)
ptr[i] = bge_vpd_readbyte(sc, i + addr);
return;
}
static void
bge_vpd_read(sc)
struct bge_softc *sc;
{
int pos = 0, i;
struct vpd_res res;
if (sc->bge_vpd_prodname != NULL)
free(sc->bge_vpd_prodname, M_DEVBUF);
if (sc->bge_vpd_readonly != NULL)
free(sc->bge_vpd_readonly, M_DEVBUF);
sc->bge_vpd_prodname = NULL;
sc->bge_vpd_readonly = NULL;
bge_vpd_read_res(sc, &res, pos);
if (res.vr_id != VPD_RES_ID) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"bad VPD resource id: expected %x got %x\n", VPD_RES_ID,
res.vr_id);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
return;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
pos += sizeof(res);
sc->bge_vpd_prodname = malloc(res.vr_len + 1, M_DEVBUF, M_NOWAIT);
for (i = 0; i < res.vr_len; i++)
sc->bge_vpd_prodname[i] = bge_vpd_readbyte(sc, i + pos);
sc->bge_vpd_prodname[i] = '\0';
pos += i;
bge_vpd_read_res(sc, &res, pos);
if (res.vr_id != VPD_RES_READ) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"bad VPD resource id: expected %x got %x\n", VPD_RES_READ,
res.vr_id);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
pos += sizeof(res);
sc->bge_vpd_readonly = malloc(res.vr_len, M_DEVBUF, M_NOWAIT);
for (i = 0; i < res.vr_len + 1; i++)
sc->bge_vpd_readonly[i] = bge_vpd_readbyte(sc, i + pos);
return;
}
Fix warnings: the bge_bpd_read*() functions are not used (#ifdef notdef)
2001-12-18 08:03:25 +00:00
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Read a byte of data stored in the EEPROM at address 'addr.' The
* BCM570x supports both the traditional bitbang interface and an
* auto access interface for reading the EEPROM. We use the auto
* access method.
*/
static u_int8_t
bge_eeprom_getbyte(sc, addr, dest)
struct bge_softc *sc;
int addr;
u_int8_t *dest;
{
int i;
u_int32_t byte = 0;
/*
* Enable use of auto EEPROM access so we can avoid
* having to use the bitbang method.
*/
BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_AUTO_EEPROM);
/* Reset the EEPROM, load the clock period. */
CSR_WRITE_4(sc, BGE_EE_ADDR,
BGE_EEADDR_RESET|BGE_EEHALFCLK(BGE_HALFCLK_384SCL));
DELAY(20);
/* Issue the read EEPROM command. */
CSR_WRITE_4(sc, BGE_EE_ADDR, BGE_EE_READCMD | addr);
/* Wait for completion */
for(i = 0; i < BGE_TIMEOUT * 10; i++) {
DELAY(10);
if (CSR_READ_4(sc, BGE_EE_ADDR) & BGE_EEADDR_DONE)
break;
}
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "EEPROM read timed out\n");
Fix bge_eeprom_getbyte() to return 1 when timeout happens. Previously it always returned 0 which means success regardless of EEPROM status. While here, add a check whether EEPROM read is successful. Submitted by: jkim
2005-12-22 02:03:57 +00:00
return(1);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* Get result. */
byte = CSR_READ_4(sc, BGE_EE_DATA);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
*dest = (byte >> ((addr % 4) * 8)) & 0xFF;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(0);
}
/*
* Read a sequence of bytes from the EEPROM.
*/
static int
bge_read_eeprom(sc, dest, off, cnt)
struct bge_softc *sc;
caddr_t dest;
int off;
int cnt;
{
int err = 0, i;
u_int8_t byte = 0;
for (i = 0; i < cnt; i++) {
err = bge_eeprom_getbyte(sc, off + i, &byte);
if (err)
break;
*(dest + i) = byte;
}
return(err ? 1 : 0);
}
static int
bge_miibus_readreg(dev, phy, reg)
device_t dev;
int phy, reg;
{
struct bge_softc *sc;
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
u_int32_t val, autopoll;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
int i;
sc = device_get_softc(dev);
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
/*
* Broadcom's own driver always assumes the internal
* PHY is at GMII address 1. On some chips, the PHY responds
* to accesses at all addresses, which could cause us to
* bogusly attach the PHY 32 times at probe type. Always
* restricting the lookup to address 1 is simpler than
* trying to figure out which chips revisions should be
* special-cased.
*/
Add support for the BCM5703x chips. I do not have one of these cards to test; however the submitter reports that this patch works with the on-board interface on the IBM x235 server. Submitted by: Jung-uk Kim <jkim@niksun.com> MFC after: 1 month
2002-09-08 19:12:02 +00:00
if (phy != 1)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
return(0);
Teach the bge driver about the BCM5701 (specifically, the 3Com 3c996B-T, with the 5701 rev B5 ASIC). One thing that confuses me still is that the 'link state change' bit in the status block seems to change state an awful lot. I have a workaround for this in place now, but it needs more investigation. For the moment though, this is enough to get the driver to work with this card.
2002-03-22 06:45:40 +00:00
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
/* Reading with autopolling on may trigger PCI errors */
autopoll = CSR_READ_4(sc, BGE_MI_MODE);
if (autopoll & BGE_MIMODE_AUTOPOLL) {
BGE_CLRBIT(sc, BGE_MI_MODE, BGE_MIMODE_AUTOPOLL);
DELAY(40);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MI_COMM, BGE_MICMD_READ|BGE_MICOMM_BUSY|
BGE_MIPHY(phy)|BGE_MIREG(reg));
for (i = 0; i < BGE_TIMEOUT; i++) {
val = CSR_READ_4(sc, BGE_MI_COMM);
if (!(val & BGE_MICOMM_BUSY))
break;
}
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
if_printf(sc->bge_ifp, "PHY read timed out\n");
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
val = 0;
goto done;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
val = CSR_READ_4(sc, BGE_MI_COMM);
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
done:
if (autopoll & BGE_MIMODE_AUTOPOLL) {
BGE_SETBIT(sc, BGE_MI_MODE, BGE_MIMODE_AUTOPOLL);
DELAY(40);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (val & BGE_MICOMM_READFAIL)
return(0);
return(val & 0xFFFF);
}
static int
bge_miibus_writereg(dev, phy, reg, val)
device_t dev;
int phy, reg, val;
{
struct bge_softc *sc;
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
u_int32_t autopoll;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
int i;
sc = device_get_softc(dev);
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
/* Reading with autopolling on may trigger PCI errors */
autopoll = CSR_READ_4(sc, BGE_MI_MODE);
if (autopoll & BGE_MIMODE_AUTOPOLL) {
BGE_CLRBIT(sc, BGE_MI_MODE, BGE_MIMODE_AUTOPOLL);
DELAY(40);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MI_COMM, BGE_MICMD_WRITE|BGE_MICOMM_BUSY|
BGE_MIPHY(phy)|BGE_MIREG(reg)|val);
for (i = 0; i < BGE_TIMEOUT; i++) {
if (!(CSR_READ_4(sc, BGE_MI_COMM) & BGE_MICOMM_BUSY))
break;
}
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
if (autopoll & BGE_MIMODE_AUTOPOLL) {
BGE_SETBIT(sc, BGE_MI_MODE, BGE_MIMODE_AUTOPOLL);
DELAY(40);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
if_printf(sc->bge_ifp, "PHY read timed out\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(0);
}
return(0);
}
static void
bge_miibus_statchg(dev)
device_t dev;
{
struct bge_softc *sc;
struct mii_data *mii;
sc = device_get_softc(dev);
mii = device_get_softc(sc->bge_miibus);
BGE_CLRBIT(sc, BGE_MAC_MODE, BGE_MACMODE_PORTMODE);
Follow NetBSD and s/IFM_1000_TX/IFM_1000_T/
2002-04-28 20:34:20 +00:00
if (IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_T) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_SETBIT(sc, BGE_MAC_MODE, BGE_PORTMODE_GMII);
} else {
BGE_SETBIT(sc, BGE_MAC_MODE, BGE_PORTMODE_MII);
}
if ((mii->mii_media_active & IFM_GMASK) == IFM_FDX) {
BGE_CLRBIT(sc, BGE_MAC_MODE, BGE_MACMODE_HALF_DUPLEX);
} else {
BGE_SETBIT(sc, BGE_MAC_MODE, BGE_MACMODE_HALF_DUPLEX);
}
return;
}
/*
* Intialize a standard receive ring descriptor.
*/
static int
bge_newbuf_std(sc, i, m)
struct bge_softc *sc;
int i;
struct mbuf *m;
{
struct mbuf *m_new = NULL;
struct bge_rx_bd *r;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
struct bge_dmamap_arg ctx;
int error;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (m == NULL) {
Use m_getcl() instead of getting mbuf, and then getting a cluster.
2006-01-23 15:57:02 +00:00
m_new = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
if (m_new == NULL)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(ENOBUFS);
m_new->m_len = m_new->m_pkthdr.len = MCLBYTES;
} else {
m_new = m;
m_new->m_len = m_new->m_pkthdr.len = MCLBYTES;
m_new->m_data = m_new->m_ext.ext_buf;
}
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
if (!sc->bge_rx_alignment_bug)
m_adj(m_new, ETHER_ALIGN);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc->bge_cdata.bge_rx_std_chain[i] = m_new;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
r = &sc->bge_ldata.bge_rx_std_ring[i];
ctx.bge_maxsegs = 1;
ctx.sc = sc;
error = bus_dmamap_load(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i], mtod(m_new, void *),
m_new->m_len, bge_dma_map_addr, &ctx, BUS_DMA_NOWAIT);
if (error || ctx.bge_maxsegs == 0) {
If bus_dmamap_load() failed, we free the mbuf. We also need to clear the pointer, to avoid double free on next bge_stop().
2005-11-30 12:37:07 +00:00
if (m == NULL) {
sc->bge_cdata.bge_rx_std_chain[i] = NULL;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
m_freem(m_new);
If bus_dmamap_load() failed, we free the mbuf. We also need to clear the pointer, to avoid double free on next bge_stop().
2005-11-30 12:37:07 +00:00
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return(ENOMEM);
}
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
r->bge_addr.bge_addr_lo = BGE_ADDR_LO(ctx.bge_busaddr);
r->bge_addr.bge_addr_hi = BGE_ADDR_HI(ctx.bge_busaddr);
r->bge_flags = BGE_RXBDFLAG_END;
r->bge_len = m_new->m_len;
r->bge_idx = i;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i],
BUS_DMASYNC_PREREAD);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(0);
}
/*
* Initialize a jumbo receive ring descriptor. This allocates
* a jumbo buffer from the pool managed internally by the driver.
*/
static int
bge_newbuf_jumbo(sc, i, m)
struct bge_softc *sc;
int i;
struct mbuf *m;
{
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
bus_dma_segment_t segs[BGE_NSEG_JUMBO];
struct bge_extrx_bd *r;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
struct mbuf *m_new = NULL;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
int nsegs;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
int error;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (m == NULL) {
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
MGETHDR(m_new, M_DONTWAIT, MT_DATA);
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
if (m_new == NULL)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(ENOBUFS);
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
m_cljget(m_new, M_DONTWAIT, MJUM9BYTES);
if (!(m_new->m_flags & M_EXT)) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m_freem(m_new);
return(ENOBUFS);
}
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
m_new->m_len = m_new->m_pkthdr.len = MJUM9BYTES;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
} else {
m_new = m;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
m_new->m_len = m_new->m_pkthdr.len = MJUM9BYTES;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m_new->m_data = m_new->m_ext.ext_buf;
}
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
if (!sc->bge_rx_alignment_bug)
m_adj(m_new, ETHER_ALIGN);
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i],
m_new, segs, &nsegs, BUS_DMA_NOWAIT);
if (error) {
if (m == NULL)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
m_freem(m_new);
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
return(error);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
}
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
sc->bge_cdata.bge_rx_jumbo_chain[i] = m_new;
/*
* Fill in the extended RX buffer descriptor.
*/
r = &sc->bge_ldata.bge_rx_jumbo_ring[i];
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
r->bge_flags = BGE_RXBDFLAG_JUMBO_RING|BGE_RXBDFLAG_END;
r->bge_idx = i;
Be ready to a case when not a constant number of segments is returned by bus_dmamap_load_mbuf_sg() on jumbo buffer allocation. Reviewed by: scottl, gallatin
2006-01-18 14:31:21 +00:00
r->bge_len3 = r->bge_len2 = r->bge_len1 = 0;
switch (nsegs) {
case 4:
r->bge_addr3.bge_addr_lo = BGE_ADDR_LO(segs[3].ds_addr);
r->bge_addr3.bge_addr_hi = BGE_ADDR_HI(segs[3].ds_addr);
r->bge_len3 = segs[3].ds_len;
case 3:
r->bge_addr2.bge_addr_lo = BGE_ADDR_LO(segs[2].ds_addr);
r->bge_addr2.bge_addr_hi = BGE_ADDR_HI(segs[2].ds_addr);
r->bge_len2 = segs[2].ds_len;
case 2:
r->bge_addr1.bge_addr_lo = BGE_ADDR_LO(segs[1].ds_addr);
r->bge_addr1.bge_addr_hi = BGE_ADDR_HI(segs[1].ds_addr);
r->bge_len1 = segs[1].ds_len;
case 1:
r->bge_addr0.bge_addr_lo = BGE_ADDR_LO(segs[0].ds_addr);
r->bge_addr0.bge_addr_hi = BGE_ADDR_HI(segs[0].ds_addr);
r->bge_len0 = segs[0].ds_len;
break;
default:
panic("%s: %d segments\n", __func__, nsegs);
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_jumbo_dmamap[i],
BUS_DMASYNC_PREREAD);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* The standard receive ring has 512 entries in it. At 2K per mbuf cluster,
* that's 1MB or memory, which is a lot. For now, we fill only the first
* 256 ring entries and hope that our CPU is fast enough to keep up with
* the NIC.
*/
static int
bge_init_rx_ring_std(sc)
struct bge_softc *sc;
{
int i;
for (i = 0; i < BGE_SSLOTS; i++) {
if (bge_newbuf_std(sc, i, NULL) == ENOBUFS)
return(ENOBUFS);
};
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map,
BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc->bge_std = i - 1;
CSR_WRITE_4(sc, BGE_MBX_RX_STD_PROD_LO, sc->bge_std);
return(0);
}
static void
bge_free_rx_ring_std(sc)
struct bge_softc *sc;
{
int i;
for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
if (sc->bge_cdata.bge_rx_std_chain[i] != NULL) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i],
BUS_DMASYNC_POSTREAD);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i]);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
m_freem(sc->bge_cdata.bge_rx_std_chain[i]);
sc->bge_cdata.bge_rx_std_chain[i] = NULL;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bzero((char *)&sc->bge_ldata.bge_rx_std_ring[i],
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sizeof(struct bge_rx_bd));
}
return;
}
static int
bge_init_rx_ring_jumbo(sc)
struct bge_softc *sc;
{
struct bge_rcb *rcb;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
int i;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
if (bge_newbuf_jumbo(sc, i, NULL) == ENOBUFS)
return(ENOBUFS);
};
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map,
BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc->bge_jumbo = i - 1;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
rcb = &sc->bge_ldata.bge_info.bge_jumbo_rx_rcb;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(0,
BGE_RCB_FLAG_USE_EXT_RX_BD);
Don't reference the maxlen and flags fields in the ring control blocks as separate 16-bit entities. Some of the ring control blocks are in NIC memory, so they must be referenced using 32-bit accesses. Smaller accesses have been observed to fail under some conditions. This caused the rings to be set up wrong, leading to writes by the card outside of the intended bounds of the rings. This problem was diagnosed by Michael Barthelow. Don Bowman submitted a patch which fixed the problem using a slightly different approach. Reference ring control blocks in NIC memory using a pointer to volatile. Parenthesize the BGE_HOSTADDR macro definition properly. MFC after: 3 days
2003-01-06 23:46:47 +00:00
CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_MAXLEN_FLAGS, rcb->bge_maxlen_flags);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MBX_RX_JUMBO_PROD_LO, sc->bge_jumbo);
return(0);
}
static void
bge_free_rx_ring_jumbo(sc)
struct bge_softc *sc;
{
int i;
for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
if (sc->bge_cdata.bge_rx_jumbo_chain[i] != NULL) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i],
BUS_DMASYNC_POSTREAD);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
m_freem(sc->bge_cdata.bge_rx_jumbo_chain[i]);
sc->bge_cdata.bge_rx_jumbo_chain[i] = NULL;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bzero((char *)&sc->bge_ldata.bge_rx_jumbo_ring[i],
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
sizeof(struct bge_extrx_bd));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
return;
}
static void
bge_free_tx_ring(sc)
struct bge_softc *sc;
{
int i;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
if (sc->bge_ldata.bge_tx_ring == NULL)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
for (i = 0; i < BGE_TX_RING_CNT; i++) {
if (sc->bge_cdata.bge_tx_chain[i] != NULL) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[i],
BUS_DMASYNC_POSTWRITE);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[i]);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
m_freem(sc->bge_cdata.bge_tx_chain[i]);
sc->bge_cdata.bge_tx_chain[i] = NULL;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bzero((char *)&sc->bge_ldata.bge_tx_ring[i],
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sizeof(struct bge_tx_bd));
}
return;
}
static int
bge_init_tx_ring(sc)
struct bge_softc *sc;
{
sc->bge_txcnt = 0;
sc->bge_tx_saved_considx = 0;
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
/* Initialize transmit producer index for host-memory send ring. */
sc->bge_tx_prodidx = 0;
CSR_WRITE_4(sc, BGE_MBX_TX_HOST_PROD0_LO, sc->bge_tx_prodidx);
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
/* 5700 b2 errata */
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
CSR_WRITE_4(sc, BGE_MBX_TX_HOST_PROD0_LO, sc->bge_tx_prodidx);
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
/* NIC-memory send ring not used; initialize to zero. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MBX_TX_NIC_PROD0_LO, 0);
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
/* 5700 b2 errata */
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
CSR_WRITE_4(sc, BGE_MBX_TX_NIC_PROD0_LO, 0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(0);
}
static void
bge_setmulti(sc)
struct bge_softc *sc;
{
struct ifnet *ifp;
struct ifmultiaddr *ifma;
u_int32_t hashes[4] = { 0, 0, 0, 0 };
int h, i;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (ifp->if_flags & IFF_ALLMULTI || ifp->if_flags & IFF_PROMISC) {
for (i = 0; i < 4; i++)
CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), 0xFFFFFFFF);
return;
}
/* First, zot all the existing filters. */
for (i = 0; i < 4; i++)
CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), 0);
/* Now program new ones. */
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
IF_ADDR_LOCK(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
Replace handrolled CRC calculation with ether_crc32_[lb]e().
2004-06-09 14:34:04 +00:00
h = ether_crc32_le(LLADDR((struct sockaddr_dl *)
ifma->ifma_addr), ETHER_ADDR_LEN) & 0x7F;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
hashes[(h & 0x60) >> 5] |= 1 << (h & 0x1F);
}
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
IF_ADDR_UNLOCK(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
for (i = 0; i < 4; i++)
CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), hashes[i]);
return;
}
/*
* Do endian, PCI and DMA initialization. Also check the on-board ROM
* self-test results.
*/
static int
bge_chipinit(sc)
struct bge_softc *sc;
{
int i;
- Change the short hand representation of the various ASIC revisions - Implement the ONEDMA_AT_ONCE workaround as described in the 5703/5704 eratta documents. Obtained from: NetBSD & Broadcom documentation
2003-05-03 22:58:45 +00:00
u_int32_t dma_rw_ctl;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
/* Set endian type before we access any non-PCI registers. */
pci_write_config(sc->bge_dev, BGE_PCI_MISC_CTL, BGE_INIT, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Check the 'ROM failed' bit on the RX CPU to see if
* self-tests passed.
*/
if (CSR_READ_4(sc, BGE_RXCPU_MODE) & BGE_RXCPUMODE_ROMFAIL) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "RX CPU self-diagnostics failed!\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(ENODEV);
}
/* Clear the MAC control register */
CSR_WRITE_4(sc, BGE_MAC_MODE, 0);
/*
* Clear the MAC statistics block in the NIC's
* internal memory.
*/
for (i = BGE_STATS_BLOCK;
i < BGE_STATS_BLOCK_END + 1; i += sizeof(u_int32_t))
BGE_MEMWIN_WRITE(sc, i, 0);
for (i = BGE_STATUS_BLOCK;
i < BGE_STATUS_BLOCK_END + 1; i += sizeof(u_int32_t))
BGE_MEMWIN_WRITE(sc, i, 0);
/* Set up the PCI DMA control register. */
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
if (sc->bge_pcie) {
dma_rw_ctl = BGE_PCI_READ_CMD|BGE_PCI_WRITE_CMD |
(0xf << BGE_PCIDMARWCTL_RD_WAT_SHIFT) |
(0x2 << BGE_PCIDMARWCTL_WR_WAT_SHIFT);
} else if (pci_read_config(sc->bge_dev, BGE_PCI_PCISTATE, 4) &
Set up the BCM570x chip's DMA Read/Write Control register more correctly. This fixes the "watchdog timeout -- resetting" errors seen on Dell 2650 systems. MFC after: 1 month
2002-09-22 18:27:29 +00:00
BGE_PCISTATE_PCI_BUSMODE) {
/* Conventional PCI bus */
- Change the short hand representation of the various ASIC revisions - Implement the ONEDMA_AT_ONCE workaround as described in the 5703/5704 eratta documents. Obtained from: NetBSD & Broadcom documentation
2003-05-03 22:58:45 +00:00
dma_rw_ctl = BGE_PCI_READ_CMD|BGE_PCI_WRITE_CMD |
(0x7 << BGE_PCIDMARWCTL_RD_WAT_SHIFT) |
(0x7 << BGE_PCIDMARWCTL_WR_WAT_SHIFT) |
(0x0F);
Set up the BCM570x chip's DMA Read/Write Control register more correctly. This fixes the "watchdog timeout -- resetting" errors seen on Dell 2650 systems. MFC after: 1 month
2002-09-22 18:27:29 +00:00
} else {
/* PCI-X bus */
- Change the short hand representation of the various ASIC revisions - Implement the ONEDMA_AT_ONCE workaround as described in the 5703/5704 eratta documents. Obtained from: NetBSD & Broadcom documentation
2003-05-03 22:58:45 +00:00
/*
* The 5704 uses a different encoding of read/write
* watermarks.
*/
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5704)
- Change the short hand representation of the various ASIC revisions - Implement the ONEDMA_AT_ONCE workaround as described in the 5703/5704 eratta documents. Obtained from: NetBSD & Broadcom documentation
2003-05-03 22:58:45 +00:00
dma_rw_ctl = BGE_PCI_READ_CMD|BGE_PCI_WRITE_CMD |
(0x7 << BGE_PCIDMARWCTL_RD_WAT_SHIFT) |
(0x3 << BGE_PCIDMARWCTL_WR_WAT_SHIFT);
else
dma_rw_ctl = BGE_PCI_READ_CMD|BGE_PCI_WRITE_CMD |
(0x3 << BGE_PCIDMARWCTL_RD_WAT_SHIFT) |
(0x3 << BGE_PCIDMARWCTL_WR_WAT_SHIFT) |
(0x0F);
/*
* 5703 and 5704 need ONEDMA_AT_ONCE as a workaround
* for hardware bugs.
*/
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5703 ||
sc->bge_asicrev == BGE_ASICREV_BCM5704) {
- Change the short hand representation of the various ASIC revisions - Implement the ONEDMA_AT_ONCE workaround as described in the 5703/5704 eratta documents. Obtained from: NetBSD & Broadcom documentation
2003-05-03 22:58:45 +00:00
u_int32_t tmp;
tmp = CSR_READ_4(sc, BGE_PCI_CLKCTL) & 0x1f;
if (tmp == 0x6 || tmp == 0x7)
dma_rw_ctl |= BGE_PCIDMARWCTL_ONEDMA_ATONCE;
}
Set up the BCM570x chip's DMA Read/Write Control register more correctly. This fixes the "watchdog timeout -- resetting" errors seen on Dell 2650 systems. MFC after: 1 month
2002-09-22 18:27:29 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5703 ||
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5704 ||
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5705 ||
sc->bge_asicrev == BGE_ASICREV_BCM5750)
- Change the short hand representation of the various ASIC revisions - Implement the ONEDMA_AT_ONCE workaround as described in the 5703/5704 eratta documents. Obtained from: NetBSD & Broadcom documentation
2003-05-03 22:58:45 +00:00
dma_rw_ctl &= ~BGE_PCIDMARWCTL_MINDMA;
pci_write_config(sc->bge_dev, BGE_PCI_DMA_RW_CTL, dma_rw_ctl, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Set up general mode register.
*/
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
CSR_WRITE_4(sc, BGE_MODE_CTL, BGE_DMA_SWAP_OPTIONS|
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_MODECTL_MAC_ATTN_INTR|BGE_MODECTL_HOST_SEND_BDS|
From the Broadcom Eratta: Setting the No_CRC bit Can Cause Data Write Errors on BCM5701/03/04 The resolution is to not set the No_CRC bit in the Mode Control Register.
2003-09-23 00:37:07 +00:00
BGE_MODECTL_TX_NO_PHDR_CSUM|BGE_MODECTL_RX_NO_PHDR_CSUM);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
Follow the lead of recent changes in the 3com Linux driver by disabling memory write invalidate unconditionally. It looks like they've decided that MWI just doesn't work with these devices. Also, remove now-irrelevant code that set PCI write boundary values based on the cache line size. MFC after: 2 weeks
2002-09-22 18:58:58 +00:00
* Disable memory write invalidate. Apparently it is not supported
* properly by these devices.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
Follow the lead of recent changes in the 3com Linux driver by disabling memory write invalidate unconditionally. It looks like they've decided that MWI just doesn't work with these devices. Also, remove now-irrelevant code that set PCI write boundary values based on the cache line size. MFC after: 2 weeks
2002-09-22 18:58:58 +00:00
PCI_CLRBIT(sc->bge_dev, BGE_PCI_CMD, PCIM_CMD_MWIEN, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#ifdef __brokenalpha__
/*
* Must insure that we do not cross an 8K (bytes) boundary
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
* for DMA reads. Our highest limit is 1K bytes. This is a
* restriction on some ALPHA platforms with early revision
* 21174 PCI chipsets, such as the AlphaPC 164lx
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
Fix a couple of longstanding typos which would have made a statement fail to compile if it weren't inside "#ifdef __brokenalpha__". Submitted by: Birger Toedtmann <birger@takatukaland.de> MFC after: 2 weeks
2002-09-22 19:27:07 +00:00
PCI_SETBIT(sc->bge_dev, BGE_PCI_DMA_RW_CTL,
BGE_PCI_READ_BNDRY_1024BYTES, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#endif
/* Set the timer prescaler (always 66Mhz) */
CSR_WRITE_4(sc, BGE_MISC_CFG, 65 << 1/*BGE_32BITTIME_66MHZ*/);
return(0);
}
static int
bge_blockinit(sc)
struct bge_softc *sc;
{
struct bge_rcb *rcb;
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
bus_size_t vrcb;
bge_hostaddr taddr;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
int i;
/*
* Initialize the memory window pointer register so that
* we can access the first 32K of internal NIC RAM. This will
* allow us to set up the TX send ring RCBs and the RX return
* ring RCBs, plus other things which live in NIC memory.
*/
CSR_WRITE_4(sc, BGE_PCI_MEMWIN_BASEADDR, 0);
The BCM5704 ASIC has a smaller mbuf space than the 5700/1/2/3. Failure to configure this correctly yields many watchdog timeouts even on lightly loaded machines. This is a common complaint from users with Dell 1750 servers with built-in dual 5704 NICs.
2003-08-10 18:04:35 +00:00
/* Note: the BCM5704 has a smaller mbuf space than other chips. */
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
/* Configure mbuf memory pool */
if (sc->bge_extram) {
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_BASEADDR,
BGE_EXT_SSRAM);
The BCM5704 ASIC has a smaller mbuf space than the 5700/1/2/3. Failure to configure this correctly yields many watchdog timeouts even on lightly loaded machines. This is a common complaint from users with Dell 1750 servers with built-in dual 5704 NICs.
2003-08-10 18:04:35 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5704)
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_LEN, 0x10000);
else
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_LEN, 0x18000);
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
} else {
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_BASEADDR,
BGE_BUFFPOOL_1);
The BCM5704 ASIC has a smaller mbuf space than the 5700/1/2/3. Failure to configure this correctly yields many watchdog timeouts even on lightly loaded machines. This is a common complaint from users with Dell 1750 servers with built-in dual 5704 NICs.
2003-08-10 18:04:35 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5704)
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_LEN, 0x10000);
else
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_LEN, 0x18000);
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
/* Configure DMA resource pool */
CSR_WRITE_4(sc, BGE_BMAN_DMA_DESCPOOL_BASEADDR,
BGE_DMA_DESCRIPTORS);
CSR_WRITE_4(sc, BGE_BMAN_DMA_DESCPOOL_LEN, 0x2000);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Configure mbuf pool watermarks */
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5705 ||
sc->bge_asicrev == BGE_ASICREV_BCM5750) {
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x0);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x10);
} else {
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x50);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x20);
}
Update the mbuf watermark settings to match the latest documentation from Broadcom. These values are strongly encouraged for proper operation. MFC after: 3 days
2003-04-26 18:26:29 +00:00
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0x60);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Configure DMA resource watermarks */
CSR_WRITE_4(sc, BGE_BMAN_DMA_DESCPOOL_LOWAT, 5);
CSR_WRITE_4(sc, BGE_BMAN_DMA_DESCPOOL_HIWAT, 10);
/* Enable buffer manager */
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_BMAN_MODE,
BGE_BMANMODE_ENABLE|BGE_BMANMODE_LOMBUF_ATTN);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
/* Poll for buffer manager start indication */
for (i = 0; i < BGE_TIMEOUT; i++) {
if (CSR_READ_4(sc, BGE_BMAN_MODE) & BGE_BMANMODE_ENABLE)
break;
DELAY(10);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"buffer manager failed to start\n");
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
return(ENXIO);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* Enable flow-through queues */
CSR_WRITE_4(sc, BGE_FTQ_RESET, 0xFFFFFFFF);
CSR_WRITE_4(sc, BGE_FTQ_RESET, 0);
/* Wait until queue initialization is complete */
for (i = 0; i < BGE_TIMEOUT; i++) {
if (CSR_READ_4(sc, BGE_FTQ_RESET) == 0)
break;
DELAY(10);
}
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "flow-through queue init failed\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(ENXIO);
}
/* Initialize the standard RX ring control block */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
rcb = &sc->bge_ldata.bge_info.bge_std_rx_rcb;
rcb->bge_hostaddr.bge_addr_lo =
BGE_ADDR_LO(sc->bge_ldata.bge_rx_std_ring_paddr);
rcb->bge_hostaddr.bge_addr_hi =
BGE_ADDR_HI(sc->bge_ldata.bge_rx_std_ring_paddr);
bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_PREREAD);
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5705 ||
sc->bge_asicrev == BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(512, 0);
else
rcb->bge_maxlen_flags =
BGE_RCB_MAXLEN_FLAGS(BGE_MAX_FRAMELEN, 0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (sc->bge_extram)
rcb->bge_nicaddr = BGE_EXT_STD_RX_RINGS;
else
rcb->bge_nicaddr = BGE_STD_RX_RINGS;
Don't reference the maxlen and flags fields in the ring control blocks as separate 16-bit entities. Some of the ring control blocks are in NIC memory, so they must be referenced using 32-bit accesses. Smaller accesses have been observed to fail under some conditions. This caused the rings to be set up wrong, leading to writes by the card outside of the intended bounds of the rings. This problem was diagnosed by Michael Barthelow. Don Bowman submitted a patch which fixed the problem using a slightly different approach. Reference ring control blocks in NIC memory using a pointer to volatile. Parenthesize the BGE_HOSTADDR macro definition properly. MFC after: 3 days
2003-01-06 23:46:47 +00:00
CSR_WRITE_4(sc, BGE_RX_STD_RCB_HADDR_HI, rcb->bge_hostaddr.bge_addr_hi);
CSR_WRITE_4(sc, BGE_RX_STD_RCB_HADDR_LO, rcb->bge_hostaddr.bge_addr_lo);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Don't reference the maxlen and flags fields in the ring control blocks as separate 16-bit entities. Some of the ring control blocks are in NIC memory, so they must be referenced using 32-bit accesses. Smaller accesses have been observed to fail under some conditions. This caused the rings to be set up wrong, leading to writes by the card outside of the intended bounds of the rings. This problem was diagnosed by Michael Barthelow. Don Bowman submitted a patch which fixed the problem using a slightly different approach. Reference ring control blocks in NIC memory using a pointer to volatile. Parenthesize the BGE_HOSTADDR macro definition properly. MFC after: 3 days
2003-01-06 23:46:47 +00:00
CSR_WRITE_4(sc, BGE_RX_STD_RCB_MAXLEN_FLAGS, rcb->bge_maxlen_flags);
CSR_WRITE_4(sc, BGE_RX_STD_RCB_NICADDR, rcb->bge_nicaddr);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Initialize the jumbo RX ring control block
* We set the 'ring disabled' bit in the flags
* field until we're actually ready to start
* using this ring (i.e. once we set the MTU
* high enough to require it).
*/
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
rcb = &sc->bge_ldata.bge_info.bge_jumbo_rx_rcb;
rcb->bge_hostaddr.bge_addr_lo =
BGE_ADDR_LO(sc->bge_ldata.bge_rx_jumbo_ring_paddr);
rcb->bge_hostaddr.bge_addr_hi =
BGE_ADDR_HI(sc->bge_ldata.bge_rx_jumbo_ring_paddr);
bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map,
BUS_DMASYNC_PREREAD);
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(0,
BGE_RCB_FLAG_USE_EXT_RX_BD|BGE_RCB_FLAG_RING_DISABLED);
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
if (sc->bge_extram)
rcb->bge_nicaddr = BGE_EXT_JUMBO_RX_RINGS;
else
rcb->bge_nicaddr = BGE_JUMBO_RX_RINGS;
CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_HADDR_HI,
rcb->bge_hostaddr.bge_addr_hi);
CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_HADDR_LO,
rcb->bge_hostaddr.bge_addr_lo);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_MAXLEN_FLAGS,
rcb->bge_maxlen_flags);
CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_NICADDR, rcb->bge_nicaddr);
/* Set up dummy disabled mini ring RCB */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
rcb = &sc->bge_ldata.bge_info.bge_mini_rx_rcb;
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
rcb->bge_maxlen_flags =
BGE_RCB_MAXLEN_FLAGS(0, BGE_RCB_FLAG_RING_DISABLED);
CSR_WRITE_4(sc, BGE_RX_MINI_RCB_MAXLEN_FLAGS,
rcb->bge_maxlen_flags);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Set the BD ring replentish thresholds. The recommended
* values are 1/8th the number of descriptors allocated to
* each ring.
*/
CSR_WRITE_4(sc, BGE_RBDI_STD_REPL_THRESH, BGE_STD_RX_RING_CNT/8);
CSR_WRITE_4(sc, BGE_RBDI_JUMBO_REPL_THRESH, BGE_JUMBO_RX_RING_CNT/8);
/*
* Disable all unused send rings by setting the 'ring disabled'
* bit in the flags field of all the TX send ring control blocks.
* These are located in NIC memory.
*/
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb = BGE_MEMWIN_START + BGE_SEND_RING_RCB;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
for (i = 0; i < BGE_TX_RINGS_EXTSSRAM_MAX; i++) {
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
BGE_RCB_MAXLEN_FLAGS(0, BGE_RCB_FLAG_RING_DISABLED));
RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0);
vrcb += sizeof(struct bge_rcb);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* Configure TX RCB 0 (we use only the first ring) */
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb = BGE_MEMWIN_START + BGE_SEND_RING_RCB;
BGE_HOSTADDR(taddr, sc->bge_ldata.bge_tx_ring_paddr);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, taddr.bge_addr_hi);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, taddr.bge_addr_lo);
RCB_WRITE_4(sc, vrcb, bge_nicaddr,
BGE_NIC_TXRING_ADDR(0, BGE_TX_RING_CNT));
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
BGE_RCB_MAXLEN_FLAGS(BGE_TX_RING_CNT, 0));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Disable all unused RX return rings */
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb = BGE_MEMWIN_START + BGE_RX_RETURN_RING_RCB;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
for (i = 0; i < BGE_RX_RINGS_MAX; i++) {
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, 0);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, 0);
RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_RCB_MAXLEN_FLAGS(sc->bge_return_ring_cnt,
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
BGE_RCB_FLAG_RING_DISABLED));
RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MBX_RX_CONS0_LO +
(i * (sizeof(u_int64_t))), 0);
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb += sizeof(struct bge_rcb);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* Initialize RX ring indexes */
CSR_WRITE_4(sc, BGE_MBX_RX_STD_PROD_LO, 0);
CSR_WRITE_4(sc, BGE_MBX_RX_JUMBO_PROD_LO, 0);
CSR_WRITE_4(sc, BGE_MBX_RX_MINI_PROD_LO, 0);
/*
* Set up RX return ring 0
* Note that the NIC address for RX return rings is 0x00000000.
* The return rings live entirely within the host, so the
* nicaddr field in the RCB isn't used.
*/
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb = BGE_MEMWIN_START + BGE_RX_RETURN_RING_RCB;
BGE_HOSTADDR(taddr, sc->bge_ldata.bge_rx_return_ring_paddr);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, taddr.bge_addr_hi);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, taddr.bge_addr_lo);
RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0x00000000);
RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
BGE_RCB_MAXLEN_FLAGS(sc->bge_return_ring_cnt, 0));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Set random backoff seed for TX */
CSR_WRITE_4(sc, BGE_TX_RANDOM_BACKOFF,
- Store pointer to the link-level address right in "struct ifnet" rather than in ifindex_table[]; all (except one) accesses are through ifp anyway. IF_LLADDR() works faster, and all (except one) ifaddr_byindex() users were converted to use ifp->if_addr. - Stop storing a (pointer to) Ethernet address in "struct arpcom", and drop the IFP2ENADDR() macro; all users have been converted to use IF_LLADDR() instead.
2005-11-11 16:04:59 +00:00
IF_LLADDR(sc->bge_ifp)[0] + IF_LLADDR(sc->bge_ifp)[1] +
IF_LLADDR(sc->bge_ifp)[2] + IF_LLADDR(sc->bge_ifp)[3] +
IF_LLADDR(sc->bge_ifp)[4] + IF_LLADDR(sc->bge_ifp)[5] +
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_TX_BACKOFF_SEED_MASK);
/* Set inter-packet gap */
CSR_WRITE_4(sc, BGE_TX_LENGTHS, 0x2620);
/*
* Specify which ring to use for packets that don't match
* any RX rules.
*/
CSR_WRITE_4(sc, BGE_RX_RULES_CFG, 0x08);
/*
* Configure number of RX lists. One interrupt distribution
* list, sixteen active lists, one bad frames class.
*/
CSR_WRITE_4(sc, BGE_RXLP_CFG, 0x181);
/* Inialize RX list placement stats mask. */
CSR_WRITE_4(sc, BGE_RXLP_STATS_ENABLE_MASK, 0x007FFFFF);
CSR_WRITE_4(sc, BGE_RXLP_STATS_CTL, 0x1);
/* Disable host coalescing until we get it set up */
CSR_WRITE_4(sc, BGE_HCC_MODE, 0x00000000);
/* Poll to make sure it's shut down. */
for (i = 0; i < BGE_TIMEOUT; i++) {
if (!(CSR_READ_4(sc, BGE_HCC_MODE) & BGE_HCCMODE_ENABLE))
break;
DELAY(10);
}
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"host coalescing engine failed to idle\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(ENXIO);
}
/* Set up host coalescing defaults */
CSR_WRITE_4(sc, BGE_HCC_RX_COAL_TICKS, sc->bge_rx_coal_ticks);
CSR_WRITE_4(sc, BGE_HCC_TX_COAL_TICKS, sc->bge_tx_coal_ticks);
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS, sc->bge_rx_max_coal_bds);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS, sc->bge_tx_max_coal_bds);
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_HCC_RX_COAL_TICKS_INT, 0);
CSR_WRITE_4(sc, BGE_HCC_TX_COAL_TICKS_INT, 0);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 0);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 0);
/* Set up address of statistics block */
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATS_ADDR_HI,
BGE_ADDR_HI(sc->bge_ldata.bge_stats_paddr));
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATS_ADDR_LO,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
BGE_ADDR_LO(sc->bge_ldata.bge_stats_paddr));
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATS_BASEADDR, BGE_STATS_BLOCK);
CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_BASEADDR, BGE_STATUS_BLOCK);
CSR_WRITE_4(sc, BGE_HCC_STATS_TICKS, sc->bge_stat_ticks);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Set up address of status block */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_ADDR_HI,
BGE_ADDR_HI(sc->bge_ldata.bge_status_block_paddr));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_ADDR_LO,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
BGE_ADDR_LO(sc->bge_ldata.bge_status_block_paddr));
sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx = 0;
sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx = 0;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Turn on host coalescing state machine */
CSR_WRITE_4(sc, BGE_HCC_MODE, BGE_HCCMODE_ENABLE);
/* Turn on RX BD completion state machine and enable attentions */
CSR_WRITE_4(sc, BGE_RBDC_MODE,
BGE_RBDCMODE_ENABLE|BGE_RBDCMODE_ATTN);
/* Turn on RX list placement state machine */
CSR_WRITE_4(sc, BGE_RXLP_MODE, BGE_RXLPMODE_ENABLE);
/* Turn on RX list selector state machine. */
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_RXLS_MODE, BGE_RXLSMODE_ENABLE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Turn on DMA, clear stats */
CSR_WRITE_4(sc, BGE_MAC_MODE, BGE_MACMODE_TXDMA_ENB|
BGE_MACMODE_RXDMA_ENB|BGE_MACMODE_RX_STATS_CLEAR|
BGE_MACMODE_TX_STATS_CLEAR|BGE_MACMODE_RX_STATS_ENB|
BGE_MACMODE_TX_STATS_ENB|BGE_MACMODE_FRMHDR_DMA_ENB|
(sc->bge_tbi ? BGE_PORTMODE_TBI : BGE_PORTMODE_MII));
/* Set misc. local control, enable interrupts on attentions */
CSR_WRITE_4(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_INTR_ONATTN);
#ifdef notdef
/* Assert GPIO pins for PHY reset */
BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_MISCIO_OUT0|
BGE_MLC_MISCIO_OUT1|BGE_MLC_MISCIO_OUT2);
BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_MISCIO_OUTEN0|
BGE_MLC_MISCIO_OUTEN1|BGE_MLC_MISCIO_OUTEN2);
#endif
/* Turn on DMA completion state machine */
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_DMAC_MODE, BGE_DMACMODE_ENABLE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Turn on write DMA state machine */
CSR_WRITE_4(sc, BGE_WDMA_MODE,
BGE_WDMAMODE_ENABLE|BGE_WDMAMODE_ALL_ATTNS);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Turn on read DMA state machine */
CSR_WRITE_4(sc, BGE_RDMA_MODE,
BGE_RDMAMODE_ENABLE|BGE_RDMAMODE_ALL_ATTNS);
/* Turn on RX data completion state machine */
CSR_WRITE_4(sc, BGE_RDC_MODE, BGE_RDCMODE_ENABLE);
/* Turn on RX BD initiator state machine */
CSR_WRITE_4(sc, BGE_RBDI_MODE, BGE_RBDIMODE_ENABLE);
/* Turn on RX data and RX BD initiator state machine */
CSR_WRITE_4(sc, BGE_RDBDI_MODE, BGE_RDBDIMODE_ENABLE);
/* Turn on Mbuf cluster free state machine */
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_MBCF_MODE, BGE_MBCFMODE_ENABLE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Turn on send BD completion state machine */
CSR_WRITE_4(sc, BGE_SBDC_MODE, BGE_SBDCMODE_ENABLE);
/* Turn on send data completion state machine */
CSR_WRITE_4(sc, BGE_SDC_MODE, BGE_SDCMODE_ENABLE);
/* Turn on send data initiator state machine */
CSR_WRITE_4(sc, BGE_SDI_MODE, BGE_SDIMODE_ENABLE);
/* Turn on send BD initiator state machine */
CSR_WRITE_4(sc, BGE_SBDI_MODE, BGE_SBDIMODE_ENABLE);
/* Turn on send BD selector state machine */
CSR_WRITE_4(sc, BGE_SRS_MODE, BGE_SRSMODE_ENABLE);
CSR_WRITE_4(sc, BGE_SDI_STATS_ENABLE_MASK, 0x007FFFFF);
CSR_WRITE_4(sc, BGE_SDI_STATS_CTL,
BGE_SDISTATSCTL_ENABLE|BGE_SDISTATSCTL_FASTER);
/* ack/clear link change events */
CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED|
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_MACSTAT_CFG_CHANGED|BGE_MACSTAT_MI_COMPLETE|
BGE_MACSTAT_LINK_CHANGED);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
CSR_WRITE_4(sc, BGE_MI_STS, 0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Enable PHY auto polling (for MII/GMII only) */
if (sc->bge_tbi) {
CSR_WRITE_4(sc, BGE_MI_STS, BGE_MISTS_LINK);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
} else {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_SETBIT(sc, BGE_MI_MODE, BGE_MIMODE_AUTOPOLL|10<<16);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
sc->bge_chipid != BGE_CHIPID_BCM5700_B1)
Try to sort out the correct way to generate async link state change interrupts. This is a bit harder than it needs to be because there's more than one way to generate link attentions, at least one of which does not work on the BCM5700, but does on the 5701. For the 5701, we can safely use the 'link changed' bit in the status block, and we enable link change attentions in the mac event register. For the 5700, we have to use MII interrupts, which require checking the MAC status register rather than the status block. This requires doing an extra register access on each interrupt which I'd prefer to avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T shows that we do in fact detect the link going up and down properly on cable insertions/disconnections. Also, avoid twiddling the autopoll enable bit in the MI mode register when doing a PHY read. I think this coupled with the other changes will stop the interrupt storms Paul Saab has been harassing me about. Manually setting the link to 100baseTX full duplex seems to work ok for me. (I'm typing over the 3c996B-T right now.) Lastly, teach the driver how to recognize a 3c996B-SX by checking the hardware config word in the EEPROM in order to detect the media. We attach 5701 fiber cards correctly now, but I haven't verified that they send/receive packets yet since I don't have a second fiber interface at home. (I know that fiber 5700 cards work, so I'm keeping my fingers crossed.)
2002-04-04 06:01:31 +00:00
CSR_WRITE_4(sc, BGE_MAC_EVT_ENB,
BGE_EVTENB_MI_INTERRUPT);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
/*
* Clear any pending link state attention.
* Otherwise some link state change events may be lost until attention
* is cleared by bge_intr() -> bge_link_upd() sequence.
* It's not necessary on newer BCM chips - perhaps enabling link
* state change attentions implies clearing pending attention.
*/
CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED|
BGE_MACSTAT_CFG_CHANGED|BGE_MACSTAT_MI_COMPLETE|
BGE_MACSTAT_LINK_CHANGED);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Enable link state change attentions. */
BGE_SETBIT(sc, BGE_MAC_EVT_ENB, BGE_EVTENB_LINK_CHANGED);
return(0);
}
/*
* Probe for a Broadcom chip. Check the PCI vendor and device IDs
* against our list and return its name if we find a match. Note
* that since the Broadcom controller contains VPD support, we
* can get the device name string from the controller itself instead
* of the compiled-in string. This is a little slow, but it guarantees
* we'll always announce the right product name.
*/
static int
bge_probe(dev)
device_t dev;
{
struct bge_type *t;
struct bge_softc *sc;
Put the ASIC revision into the device description string so it appears in the dmesg output. MFC after: 3 days
2003-01-07 01:33:03 +00:00
char *descbuf;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
t = bge_devs;
sc = device_get_softc(dev);
bzero(sc, sizeof(struct bge_softc));
sc->bge_dev = dev;
while(t->bge_name != NULL) {
if ((pci_get_vendor(dev) == t->bge_vid) &&
(pci_get_device(dev) == t->bge_did)) {
#ifdef notdef
bge_vpd_read(sc);
device_set_desc(dev, sc->bge_vpd_prodname);
#endif
Put the ASIC revision into the device description string so it appears in the dmesg output. MFC after: 3 days
2003-01-07 01:33:03 +00:00
descbuf = malloc(BGE_DEVDESC_MAX, M_TEMP, M_NOWAIT);
if (descbuf == NULL)
return(ENOMEM);
snprintf(descbuf, BGE_DEVDESC_MAX,
"%s, ASIC rev. %#04x", t->bge_name,
pci_read_config(dev, BGE_PCI_MISC_CTL, 4) >> 16);
device_set_desc_copy(dev, descbuf);
On Dell boxes such as the PE2650 we need to disable 3 led mode for the link LED to work. Reviewed by: wpaul Obtained from: Linux driver
2003-08-20 04:06:00 +00:00
if (pci_get_subvendor(dev) == DELL_VENDORID)
sc->bge_no_3_led = 1;
Put the ASIC revision into the device description string so it appears in the dmesg output. MFC after: 3 days
2003-01-07 01:33:03 +00:00
free(descbuf, M_TEMP);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(0);
}
t++;
}
return(ENXIO);
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
static void
bge_dma_free(sc)
struct bge_softc *sc;
{
int i;
/* Destroy DMA maps for RX buffers */
for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
if (sc->bge_cdata.bge_rx_std_dmamap[i])
bus_dmamap_destroy(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i]);
}
/* Destroy DMA maps for jumbo RX buffers */
for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
if (sc->bge_cdata.bge_rx_jumbo_dmamap[i])
bus_dmamap_destroy(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
}
/* Destroy DMA maps for TX buffers */
for (i = 0; i < BGE_TX_RING_CNT; i++) {
if (sc->bge_cdata.bge_tx_dmamap[i])
bus_dmamap_destroy(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[i]);
}
if (sc->bge_cdata.bge_mtag)
bus_dma_tag_destroy(sc->bge_cdata.bge_mtag);
/* Destroy standard RX ring */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_std_ring_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_std_ring_map && sc->bge_ldata.bge_rx_std_ring)
bus_dmamem_free(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_ldata.bge_rx_std_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_rx_std_ring_map);
if (sc->bge_cdata.bge_rx_std_ring_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_rx_std_ring_tag);
/* Destroy jumbo RX ring */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_jumbo_ring_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_jumbo_ring_map &&
sc->bge_ldata.bge_rx_jumbo_ring)
bus_dmamem_free(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_ldata.bge_rx_jumbo_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_rx_jumbo_ring_map);
if (sc->bge_cdata.bge_rx_jumbo_ring_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_rx_jumbo_ring_tag);
/* Destroy RX return ring */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_return_ring_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_rx_return_ring_tag,
sc->bge_cdata.bge_rx_return_ring_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_return_ring_map &&
sc->bge_ldata.bge_rx_return_ring)
bus_dmamem_free(sc->bge_cdata.bge_rx_return_ring_tag,
sc->bge_ldata.bge_rx_return_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_rx_return_ring_map);
if (sc->bge_cdata.bge_rx_return_ring_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_rx_return_ring_tag);
/* Destroy TX ring */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_tx_ring_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_cdata.bge_tx_ring_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_tx_ring_map && sc->bge_ldata.bge_tx_ring)
bus_dmamem_free(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_ldata.bge_tx_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_tx_ring_map);
if (sc->bge_cdata.bge_tx_ring_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_tx_ring_tag);
/* Destroy status block */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_status_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_status_map && sc->bge_ldata.bge_status_block)
bus_dmamem_free(sc->bge_cdata.bge_status_tag,
sc->bge_ldata.bge_status_block,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_status_map);
if (sc->bge_cdata.bge_status_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_status_tag);
/* Destroy statistics block */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_stats_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_stats_tag,
sc->bge_cdata.bge_stats_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_stats_map && sc->bge_ldata.bge_stats)
bus_dmamem_free(sc->bge_cdata.bge_stats_tag,
sc->bge_ldata.bge_stats,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_stats_map);
if (sc->bge_cdata.bge_stats_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_stats_tag);
/* Destroy the parent tag */
if (sc->bge_cdata.bge_parent_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_parent_tag);
return;
}
static int
bge_dma_alloc(dev)
device_t dev;
{
struct bge_softc *sc;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
int i, error;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
struct bge_dmamap_arg ctx;
sc = device_get_softc(dev);
/*
* Allocate the parent bus DMA tag appropriate for PCI.
*/
error = bus_dma_tag_create(NULL, /* parent */
PAGE_SIZE, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
Fix the order of the lowaddr,highaddr arguments in the parent tag. This coincidentally didn't cause any problems, but was definitely wrong.
2005-04-24 02:45:27 +00:00
BUS_SPACE_MAXADDR, /* highaddr */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
NULL, NULL, /* filter, filterarg */
MAXBSIZE, BGE_NSEG_NEW, /* maxsize, nsegments */
BUS_SPACE_MAXSIZE_32BIT,/* maxsegsize */
Don't force busdma to pre-allocate bounce pages for the parent tag.
2004-11-21 04:02:36 +00:00
0, /* flags */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
NULL, NULL, /* lockfunc, lockarg */
&sc->bge_cdata.bge_parent_tag);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (error != 0) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"could not allocate parent dma tag\n");
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
return (ENOMEM);
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/*
* Create tag for RX mbufs.
*/
Use and alignment of 1 instead of ETHER_ALIGN for rx and tx buffers and jumbo frames. BGE hardware with the rx alignment bug will still be handled by the calls to m_adj() that already exist. m_adj() is probably better suited for this task anyways. Just as with if_em, this saves a malloc + several locks per packet and prevents unneeded data copying within busdma.
2004-10-19 02:42:49 +00:00
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag, 1,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
NULL, MCLBYTES * BGE_NSEG_NEW, BGE_NSEG_NEW, MCLBYTES,
BUS_DMA_ALLOCNOW, NULL, NULL, &sc->bge_cdata.bge_mtag);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
/* Create DMA maps for RX buffers */
for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
error = bus_dmamap_create(sc->bge_cdata.bge_mtag, 0,
&sc->bge_cdata.bge_rx_std_dmamap[i]);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"can't create DMA map for RX\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return(ENOMEM);
}
}
/* Create DMA maps for TX buffers */
for (i = 0; i < BGE_TX_RING_CNT; i++) {
error = bus_dmamap_create(sc->bge_cdata.bge_mtag, 0,
&sc->bge_cdata.bge_tx_dmamap[i]);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"can't create DMA map for RX\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return(ENOMEM);
}
}
/* Create tag for standard RX ring */
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_STD_RX_RING_SZ, 1, BGE_STD_RX_RING_SZ, 0,
NULL, NULL, &sc->bge_cdata.bge_rx_std_ring_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
/* Allocate DMA'able memory for standard RX ring */
error = bus_dmamem_alloc(sc->bge_cdata.bge_rx_std_ring_tag,
(void **)&sc->bge_ldata.bge_rx_std_ring, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_rx_std_ring_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_rx_std_ring, BGE_STD_RX_RING_SZ);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/* Load the address of the standard RX ring */
ctx.bge_maxsegs = 1;
ctx.sc = sc;
error = bus_dmamap_load(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map, sc->bge_ldata.bge_rx_std_ring,
BGE_STD_RX_RING_SZ, bge_dma_map_addr, &ctx, BUS_DMA_NOWAIT);
if (error)
return (ENOMEM);
sc->bge_ldata.bge_rx_std_ring_paddr = ctx.bge_busaddr;
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/*
* Create tag for jumbo mbufs.
* This is really a bit of a kludge. We allocate a special
* jumbo buffer pool which (thanks to the way our DMA
* memory allocation works) will consist of contiguous
* pages. This means that even though a jumbo buffer might
* be larger than a page size, we don't really need to
* map it into more than one DMA segment. However, the
* default mbuf tag will result in multi-segment mappings,
* so we have to create a special jumbo mbuf tag that
* lets us get away with mapping the jumbo buffers as
* a single segment. I think eventually the driver should
* be changed so that it uses ordinary mbufs and cluster
* buffers, i.e. jumbo frames can span multiple DMA
* descriptors. But that's a project for another day.
*/
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
Use and alignment of 1 instead of ETHER_ALIGN for rx and tx buffers and jumbo frames. BGE hardware with the rx alignment bug will still be handled by the calls to m_adj() that already exist. m_adj() is probably better suited for this task anyways. Just as with if_em, this saves a malloc + several locks per packet and prevents unneeded data copying within busdma.
2004-10-19 02:42:49 +00:00
1, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
NULL, MJUM9BYTES, BGE_NSEG_JUMBO, PAGE_SIZE,
0, NULL, NULL, &sc->bge_cdata.bge_mtag_jumbo);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
/* Create tag for jumbo RX ring */
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_JUMBO_RX_RING_SZ, 1, BGE_JUMBO_RX_RING_SZ, 0,
NULL, NULL, &sc->bge_cdata.bge_rx_jumbo_ring_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
/* Allocate DMA'able memory for jumbo RX ring */
error = bus_dmamem_alloc(sc->bge_cdata.bge_rx_jumbo_ring_tag,
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
(void **)&sc->bge_ldata.bge_rx_jumbo_ring,
BUS_DMA_NOWAIT | BUS_DMA_ZERO,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
&sc->bge_cdata.bge_rx_jumbo_ring_map);
if (error)
return (ENOMEM);
/* Load the address of the jumbo RX ring */
ctx.bge_maxsegs = 1;
ctx.sc = sc;
error = bus_dmamap_load(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map,
sc->bge_ldata.bge_rx_jumbo_ring, BGE_JUMBO_RX_RING_SZ,
bge_dma_map_addr, &ctx, BUS_DMA_NOWAIT);
if (error)
return (ENOMEM);
sc->bge_ldata.bge_rx_jumbo_ring_paddr = ctx.bge_busaddr;
/* Create DMA maps for jumbo RX buffers */
for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
error = bus_dmamap_create(sc->bge_cdata.bge_mtag_jumbo,
0, &sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
"can't create DMA map for RX\n");
return(ENOMEM);
}
}
}
/* Create tag for RX return ring */
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_RX_RTN_RING_SZ(sc), 1, BGE_RX_RTN_RING_SZ(sc), 0,
NULL, NULL, &sc->bge_cdata.bge_rx_return_ring_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
/* Allocate DMA'able memory for RX return ring */
error = bus_dmamem_alloc(sc->bge_cdata.bge_rx_return_ring_tag,
(void **)&sc->bge_ldata.bge_rx_return_ring, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_rx_return_ring_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_rx_return_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
BGE_RX_RTN_RING_SZ(sc));
/* Load the address of the RX return ring */
ctx.bge_maxsegs = 1;
ctx.sc = sc;
error = bus_dmamap_load(sc->bge_cdata.bge_rx_return_ring_tag,
sc->bge_cdata.bge_rx_return_ring_map,
sc->bge_ldata.bge_rx_return_ring, BGE_RX_RTN_RING_SZ(sc),
bge_dma_map_addr, &ctx, BUS_DMA_NOWAIT);
if (error)
return (ENOMEM);
sc->bge_ldata.bge_rx_return_ring_paddr = ctx.bge_busaddr;
/* Create tag for TX ring */
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_TX_RING_SZ, 1, BGE_TX_RING_SZ, 0, NULL, NULL,
&sc->bge_cdata.bge_tx_ring_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
/* Allocate DMA'able memory for TX ring */
error = bus_dmamem_alloc(sc->bge_cdata.bge_tx_ring_tag,
(void **)&sc->bge_ldata.bge_tx_ring, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_tx_ring_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_tx_ring, BGE_TX_RING_SZ);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/* Load the address of the TX ring */
ctx.bge_maxsegs = 1;
ctx.sc = sc;
error = bus_dmamap_load(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_cdata.bge_tx_ring_map, sc->bge_ldata.bge_tx_ring,
BGE_TX_RING_SZ, bge_dma_map_addr, &ctx, BUS_DMA_NOWAIT);
if (error)
return (ENOMEM);
sc->bge_ldata.bge_tx_ring_paddr = ctx.bge_busaddr;
/* Create tag for status block */
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_STATUS_BLK_SZ, 1, BGE_STATUS_BLK_SZ, 0,
NULL, NULL, &sc->bge_cdata.bge_status_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
/* Allocate DMA'able memory for status block */
error = bus_dmamem_alloc(sc->bge_cdata.bge_status_tag,
(void **)&sc->bge_ldata.bge_status_block, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_status_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_status_block, BGE_STATUS_BLK_SZ);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/* Load the address of the status block */
ctx.sc = sc;
ctx.bge_maxsegs = 1;
error = bus_dmamap_load(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, sc->bge_ldata.bge_status_block,
BGE_STATUS_BLK_SZ, bge_dma_map_addr, &ctx, BUS_DMA_NOWAIT);
if (error)
return (ENOMEM);
sc->bge_ldata.bge_status_block_paddr = ctx.bge_busaddr;
/* Create tag for statistics block */
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
NULL, BGE_STATS_SZ, 1, BGE_STATS_SZ, 0, NULL, NULL,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
&sc->bge_cdata.bge_stats_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
/* Allocate DMA'able memory for statistics block */
error = bus_dmamem_alloc(sc->bge_cdata.bge_stats_tag,
(void **)&sc->bge_ldata.bge_stats, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_stats_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_stats, BGE_STATS_SZ);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/* Load the address of the statstics block */
ctx.sc = sc;
ctx.bge_maxsegs = 1;
error = bus_dmamap_load(sc->bge_cdata.bge_stats_tag,
sc->bge_cdata.bge_stats_map, sc->bge_ldata.bge_stats,
BGE_STATS_SZ, bge_dma_map_addr, &ctx, BUS_DMA_NOWAIT);
if (error)
return (ENOMEM);
sc->bge_ldata.bge_stats_paddr = ctx.bge_busaddr;
return(0);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static int
bge_attach(dev)
device_t dev;
{
struct ifnet *ifp;
struct bge_softc *sc;
Try to sort out the correct way to generate async link state change interrupts. This is a bit harder than it needs to be because there's more than one way to generate link attentions, at least one of which does not work on the BCM5700, but does on the 5701. For the 5701, we can safely use the 'link changed' bit in the status block, and we enable link change attentions in the mac event register. For the 5700, we have to use MII interrupts, which require checking the MAC status register rather than the status block. This requires doing an extra register access on each interrupt which I'd prefer to avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T shows that we do in fact detect the link going up and down properly on cable insertions/disconnections. Also, avoid twiddling the autopoll enable bit in the MI mode register when doing a PHY read. I think this coupled with the other changes will stop the interrupt storms Paul Saab has been harassing me about. Manually setting the link to 100baseTX full duplex seems to work ok for me. (I'm typing over the 3c996B-T right now.) Lastly, teach the driver how to recognize a 3c996B-SX by checking the hardware config word in the EEPROM in order to detect the media. We attach 5701 fiber cards correctly now, but I haven't verified that they send/receive packets yet since I don't have a second fiber interface at home. (I know that fiber 5700 cards work, so I'm keeping my fingers crossed.)
2002-04-04 06:01:31 +00:00
u_int32_t hwcfg = 0;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
u_int32_t mac_tmp = 0;
u_char eaddr[6];
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
int error = 0, rid;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc = device_get_softc(dev);
sc->bge_dev = dev;
/*
* Map control/status registers.
*/
pci_enable_busmaster(dev);
rid = BGE_PCI_BAR0;
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->bge_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
RF_ACTIVE|PCI_RF_DENSE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (sc->bge_res == NULL) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf (sc->bge_dev, "couldn't map memory\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = ENXIO;
goto fail;
}
sc->bge_btag = rman_get_bustag(sc->bge_res);
sc->bge_bhandle = rman_get_bushandle(sc->bge_res);
/* Allocate interrupt */
rid = 0;
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
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->bge_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
RF_SHAREABLE | RF_ACTIVE);
if (sc->bge_irq == NULL) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "couldn't map interrupt\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = ENXIO;
goto fail;
}
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_INIT(sc, device_get_nameunit(dev));
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
/* Save ASIC rev. */
sc->bge_chipid =
pci_read_config(dev, BGE_PCI_MISC_CTL, 4) &
BGE_PCIMISCCTL_ASICREV;
sc->bge_asicrev = BGE_ASICREV(sc->bge_chipid);
sc->bge_chiprev = BGE_CHIPREV(sc->bge_chipid);
Support the 5714C Submitted by: John Cagle <john dot cagle at hp dot com>
2005-05-19 21:08:59 +00:00
/*
Recognize Broadcom BCM5752 chip, that can be found in HP DC7600. PR: kern/88940 Submitted by: Alexander Hausner
2005-11-15 14:43:22 +00:00
* Treat the 5714 and the 5752 like the 5750 until we have more info
Support the 5714C Submitted by: John Cagle <john dot cagle at hp dot com>
2005-05-19 21:08:59 +00:00
* on this chip.
*/
Recognize Broadcom BCM5752 chip, that can be found in HP DC7600. PR: kern/88940 Submitted by: Alexander Hausner
2005-11-15 14:43:22 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5714 ||
sc->bge_asicrev == BGE_ASICREV_BCM5752)
Support the 5714C Submitted by: John Cagle <john dot cagle at hp dot com>
2005-05-19 21:08:59 +00:00
sc->bge_asicrev = BGE_ASICREV_BCM5750;
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
/*
* XXX: Broadcom Linux driver. Not in specs or eratta.
* PCI-Express?
*/
if (sc->bge_asicrev == BGE_ASICREV_BCM5750) {
u_int32_t v;
v = pci_read_config(dev, BGE_PCI_MSI_CAPID, 4);
if (((v >> 8) & 0xff) == BGE_PCIE_CAPID_REG) {
v = pci_read_config(dev, BGE_PCIE_CAPID_REG, 4);
if ((v & 0xff) == BGE_PCIE_CAPID)
sc->bge_pcie = 1;
}
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Try to reset the chip. */
bge_reset(sc);
if (bge_chipinit(sc)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "chip initialization failed\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
/*
* Get station address from the EEPROM.
*/
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
mac_tmp = bge_readmem_ind(sc, 0x0c14);
if ((mac_tmp >> 16) == 0x484b) {
eaddr[0] = (u_char)(mac_tmp >> 8);
eaddr[1] = (u_char)mac_tmp;
mac_tmp = bge_readmem_ind(sc, 0x0c18);
eaddr[2] = (u_char)(mac_tmp >> 24);
eaddr[3] = (u_char)(mac_tmp >> 16);
eaddr[4] = (u_char)(mac_tmp >> 8);
eaddr[5] = (u_char)mac_tmp;
} else if (bge_read_eeprom(sc, eaddr,
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_EE_MAC_OFFSET + 2, ETHER_ADDR_LEN)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "failed to read station address\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/* 5705 limits RX return ring to 512 entries. */
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5705 ||
sc->bge_asicrev == BGE_ASICREV_BCM5750)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_return_ring_cnt = BGE_RETURN_RING_CNT_5705;
else
sc->bge_return_ring_cnt = BGE_RETURN_RING_CNT;
if (bge_dma_alloc(dev)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"failed to allocate DMA resources\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Set default tuneable values. */
sc->bge_stat_ticks = BGE_TICKS_PER_SEC;
sc->bge_rx_coal_ticks = 150;
sc->bge_tx_coal_ticks = 150;
sc->bge_rx_max_coal_bds = 64;
sc->bge_tx_max_coal_bds = 128;
/* Set up ifnet structure */
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp = if_alloc(IFT_ETHER);
if (ifp == NULL) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "failed to if_alloc()\n");
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_softc = sc;
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));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
ifp->if_ioctl = bge_ioctl;
ifp->if_start = bge_start;
ifp->if_watchdog = bge_watchdog;
ifp->if_init = bge_init;
ifp->if_mtu = ETHERMTU;
Add altq(4) support.
2004-10-30 21:21:10 +00:00
ifp->if_snd.ifq_drv_maxlen = BGE_TX_RING_CNT - 1;
IFQ_SET_MAXLEN(&ifp->if_snd, ifp->if_snd.ifq_drv_maxlen);
IFQ_SET_READY(&ifp->if_snd);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_hwassist = BGE_CSUM_FEATURES;
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
ifp->if_capabilities = IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING |
This driver can do hardware VLAN tagging + checksum offloading. In collaboration with: Mihail Balikov <mihail.balikov interbgc.com>
2006-01-30 13:45:55 +00:00
IFCAP_VLAN_MTU | IFCAP_VLAN_HWCSUM;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_capenable = ifp->if_capabilities;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
ifp->if_capabilities |= IFCAP_POLLING;
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
/*
* 5700 B0 chips do not support checksumming correctly due
* to hardware bugs.
*/
if (sc->bge_chipid == BGE_CHIPID_BCM5700_B0) {
ifp->if_capabilities &= ~IFCAP_HWCSUM;
ifp->if_capenable &= IFCAP_HWCSUM;
ifp->if_hwassist = 0;
}
Try to sort out the correct way to generate async link state change interrupts. This is a bit harder than it needs to be because there's more than one way to generate link attentions, at least one of which does not work on the BCM5700, but does on the 5701. For the 5701, we can safely use the 'link changed' bit in the status block, and we enable link change attentions in the mac event register. For the 5700, we have to use MII interrupts, which require checking the MAC status register rather than the status block. This requires doing an extra register access on each interrupt which I'd prefer to avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T shows that we do in fact detect the link going up and down properly on cable insertions/disconnections. Also, avoid twiddling the autopoll enable bit in the MI mode register when doing a PHY read. I think this coupled with the other changes will stop the interrupt storms Paul Saab has been harassing me about. Manually setting the link to 100baseTX full duplex seems to work ok for me. (I'm typing over the 3c996B-T right now.) Lastly, teach the driver how to recognize a 3c996B-SX by checking the hardware config word in the EEPROM in order to detect the media. We attach 5701 fiber cards correctly now, but I haven't verified that they send/receive packets yet since I don't have a second fiber interface at home. (I know that fiber 5700 cards work, so I'm keeping my fingers crossed.)
2002-04-04 06:01:31 +00:00
/*
* Figure out what sort of media we have by checking the
Obtain the media type from the shared memory and only use the eeprom as a fallback.
2003-02-05 08:54:36 +00:00
* hardware config word in the first 32k of NIC internal memory,
* or fall back to examining the EEPROM if necessary.
* Note: on some BCM5700 cards, this value appears to be unset.
* If that's the case, we have to rely on identifying the NIC
* by its PCI subsystem ID, as we do below for the SysKonnect
* SK-9D41.
Try to sort out the correct way to generate async link state change interrupts. This is a bit harder than it needs to be because there's more than one way to generate link attentions, at least one of which does not work on the BCM5700, but does on the 5701. For the 5701, we can safely use the 'link changed' bit in the status block, and we enable link change attentions in the mac event register. For the 5700, we have to use MII interrupts, which require checking the MAC status register rather than the status block. This requires doing an extra register access on each interrupt which I'd prefer to avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T shows that we do in fact detect the link going up and down properly on cable insertions/disconnections. Also, avoid twiddling the autopoll enable bit in the MI mode register when doing a PHY read. I think this coupled with the other changes will stop the interrupt storms Paul Saab has been harassing me about. Manually setting the link to 100baseTX full duplex seems to work ok for me. (I'm typing over the 3c996B-T right now.) Lastly, teach the driver how to recognize a 3c996B-SX by checking the hardware config word in the EEPROM in order to detect the media. We attach 5701 fiber cards correctly now, but I haven't verified that they send/receive packets yet since I don't have a second fiber interface at home. (I know that fiber 5700 cards work, so I'm keeping my fingers crossed.)
2002-04-04 06:01:31 +00:00
*/
Obtain the media type from the shared memory and only use the eeprom as a fallback.
2003-02-05 08:54:36 +00:00
if (bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM_SIG) == BGE_MAGIC_NUMBER)
hwcfg = bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM_NICCFG);
else {
Fix bge_eeprom_getbyte() to return 1 when timeout happens. Previously it always returned 0 which means success regardless of EEPROM status. While here, add a check whether EEPROM read is successful. Submitted by: jkim
2005-12-22 02:03:57 +00:00
if (bge_read_eeprom(sc, (caddr_t)&hwcfg, BGE_EE_HWCFG_OFFSET,
sizeof(hwcfg))) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "failed to read EEPROM\n");
Fix bge_eeprom_getbyte() to return 1 when timeout happens. Previously it always returned 0 which means success regardless of EEPROM status. While here, add a check whether EEPROM read is successful. Submitted by: jkim
2005-12-22 02:03:57 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Obtain the media type from the shared memory and only use the eeprom as a fallback.
2003-02-05 08:54:36 +00:00
hwcfg = ntohl(hwcfg);
}
if ((hwcfg & BGE_HWCFG_MEDIA) == BGE_MEDIA_FIBER)
Try to sort out the correct way to generate async link state change interrupts. This is a bit harder than it needs to be because there's more than one way to generate link attentions, at least one of which does not work on the BCM5700, but does on the 5701. For the 5701, we can safely use the 'link changed' bit in the status block, and we enable link change attentions in the mac event register. For the 5700, we have to use MII interrupts, which require checking the MAC status register rather than the status block. This requires doing an extra register access on each interrupt which I'd prefer to avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T shows that we do in fact detect the link going up and down properly on cable insertions/disconnections. Also, avoid twiddling the autopoll enable bit in the MI mode register when doing a PHY read. I think this coupled with the other changes will stop the interrupt storms Paul Saab has been harassing me about. Manually setting the link to 100baseTX full duplex seems to work ok for me. (I'm typing over the 3c996B-T right now.) Lastly, teach the driver how to recognize a 3c996B-SX by checking the hardware config word in the EEPROM in order to detect the media. We attach 5701 fiber cards correctly now, but I haven't verified that they send/receive packets yet since I don't have a second fiber interface at home. (I know that fiber 5700 cards work, so I'm keeping my fingers crossed.)
2002-04-04 06:01:31 +00:00
sc->bge_tbi = 1;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* The SysKonnect SK-9D41 is a 1000baseSX card. */
if ((pci_read_config(dev, BGE_PCI_SUBSYS, 4) >> 16) == SK_SUBSYSID_9D41)
sc->bge_tbi = 1;
if (sc->bge_tbi) {
ifmedia_init(&sc->bge_ifmedia, IFM_IMASK,
bge_ifmedia_upd, bge_ifmedia_sts);
ifmedia_add(&sc->bge_ifmedia, IFM_ETHER|IFM_1000_SX, 0, NULL);
ifmedia_add(&sc->bge_ifmedia,
IFM_ETHER|IFM_1000_SX|IFM_FDX, 0, NULL);
ifmedia_add(&sc->bge_ifmedia, IFM_ETHER|IFM_AUTO, 0, NULL);
ifmedia_set(&sc->bge_ifmedia, IFM_ETHER|IFM_AUTO);
Add some special case code to fix a problem with the BCM5704 in TBI (fiber) mode. The 5704 apparently has some s00p3r s33kr1t registers for setting the advertisement of pause frame ability (i.e flow control) when in autoneg mode. If we don't set these registers correctly, we may not be able to negotiate a proper link with some switches. (Symptom is that the NIC reports the link as up (PCS synched) but no traffic can be exchanged.) PR: kern/67598
2004-06-09 16:01:59 +00:00
sc->bge_ifmedia.ifm_media = sc->bge_ifmedia.ifm_cur->ifm_media;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
} else {
/*
* Do transceiver setup.
*/
if (mii_phy_probe(dev, &sc->bge_miibus,
bge_ifmedia_upd, bge_ifmedia_sts)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "MII without any PHY!\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
}
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
/*
* When using the BCM5701 in PCI-X mode, data corruption has
* been observed in the first few bytes of some received packets.
* Aligning the packet buffer in memory eliminates the corruption.
* Unfortunately, this misaligns the packet payloads. On platforms
* which do not support unaligned accesses, we will realign the
* payloads by copying the received packets.
*/
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
switch (sc->bge_chipid) {
case BGE_CHIPID_BCM5701_A0:
case BGE_CHIPID_BCM5701_B0:
case BGE_CHIPID_BCM5701_B2:
case BGE_CHIPID_BCM5701_B5:
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
/* If in PCI-X mode, work around the alignment bug. */
if ((pci_read_config(dev, BGE_PCI_PCISTATE, 4) &
(BGE_PCISTATE_PCI_BUSMODE | BGE_PCISTATE_PCI_BUSSPEED)) ==
BGE_PCISTATE_PCI_BUSSPEED)
sc->bge_rx_alignment_bug = 1;
break;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Call MI attach routine.
*/
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ether_ifattach(ifp, eaddr);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
callout_init(&sc->bge_stat_ch, CALLOUT_MPSAFE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
/*
* Hookup IRQ last.
*/
error = bus_setup_intr(dev, sc->bge_irq, INTR_TYPE_NET | INTR_MPSAFE,
bge_intr, sc, &sc->bge_intrhand);
if (error) {
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
bge_detach(dev);
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "couldn't set up irq\n");
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
fail:
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return(error);
}
static int
bge_detach(dev)
device_t dev;
{
struct bge_softc *sc;
struct ifnet *ifp;
sc = device_get_softc(dev);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING)
ether_poll_deregister(ifp);
#endif
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_stop(sc);
bge_reset(sc);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
ether_ifdetach(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (sc->bge_tbi) {
ifmedia_removeall(&sc->bge_ifmedia);
} else {
bus_generic_detach(dev);
device_delete_child(dev, sc->bge_miibus);
}
bge_release_resources(sc);
return(0);
}
static void
bge_release_resources(sc)
struct bge_softc *sc;
{
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
device_t dev;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
dev = sc->bge_dev;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (sc->bge_vpd_prodname != NULL)
free(sc->bge_vpd_prodname, M_DEVBUF);
if (sc->bge_vpd_readonly != NULL)
free(sc->bge_vpd_readonly, M_DEVBUF);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (sc->bge_intrhand != NULL)
bus_teardown_intr(dev, sc->bge_irq, sc->bge_intrhand);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (sc->bge_irq != NULL)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bus_release_resource(dev, SYS_RES_IRQ, 0, sc->bge_irq);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (sc->bge_res != NULL)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bus_release_resource(dev, SYS_RES_MEMORY,
BGE_PCI_BAR0, sc->bge_res);
Fix "struct ifnet" leaks when attach() fails in the middle.
2005-09-16 11:25:19 +00:00
if (sc->bge_ifp != NULL)
if_free(sc->bge_ifp);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bge_dma_free(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
if (mtx_initialized(&sc->bge_mtx)) /* XXX */
BGE_LOCK_DESTROY(sc);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
return;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
static void
bge_reset(sc)
struct bge_softc *sc;
{
device_t dev;
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
u_int32_t cachesize, command, pcistate, reset;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
int i, val = 0;
dev = sc->bge_dev;
/* Save some important PCI state. */
cachesize = pci_read_config(dev, BGE_PCI_CACHESZ, 4);
command = pci_read_config(dev, BGE_PCI_CMD, 4);
pcistate = pci_read_config(dev, BGE_PCI_PCISTATE, 4);
pci_write_config(dev, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_INDIRECT_ACCESS|BGE_PCIMISCCTL_MASK_PCI_INTR|
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
BGE_HIF_SWAP_OPTIONS|BGE_PCIMISCCTL_PCISTATE_RW, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
reset = BGE_MISCCFG_RESET_CORE_CLOCKS|(65<<1);
/* XXX: Broadcom Linux driver. */
if (sc->bge_pcie) {
if (CSR_READ_4(sc, 0x7e2c) == 0x60) /* PCIE 1.0 */
CSR_WRITE_4(sc, 0x7e2c, 0x20);
if (sc->bge_chipid != BGE_CHIPID_BCM5750_A0) {
/* Prevent PCIE link training during global reset */
CSR_WRITE_4(sc, BGE_MISC_CFG, (1<<29));
reset |= (1<<29);
}
}
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Issue global reset */
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
bge_writereg_ind(sc, BGE_MISC_CFG, reset);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
DELAY(1000);
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
/* XXX: Broadcom Linux driver. */
if (sc->bge_pcie) {
if (sc->bge_chipid == BGE_CHIPID_BCM5750_A0) {
uint32_t v;
DELAY(500000); /* wait for link training to complete */
v = pci_read_config(dev, 0xc4, 4);
pci_write_config(dev, 0xc4, v | (1<<15), 4);
}
/* Set PCIE max payload size and clear error status. */
pci_write_config(dev, 0xd8, 0xf5000, 4);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Reset some of the PCI state that got zapped by reset */
pci_write_config(dev, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_INDIRECT_ACCESS|BGE_PCIMISCCTL_MASK_PCI_INTR|
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
BGE_HIF_SWAP_OPTIONS|BGE_PCIMISCCTL_PCISTATE_RW, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
pci_write_config(dev, BGE_PCI_CACHESZ, cachesize, 4);
pci_write_config(dev, BGE_PCI_CMD, command, 4);
bge_writereg_ind(sc, BGE_MISC_CFG, (65 << 1));
Enable the memory arbiter before turning off the PXE restart. This prevents NMI's from happening when resetting the chip on some hardware I have seen. Mis-behaving box made available by: John Cagle <john.cagle@hp.com>
2004-04-06 18:28:15 +00:00
/* Enable memory arbiter. */
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Enable the memory arbiter before turning off the PXE restart. This prevents NMI's from happening when resetting the chip on some hardware I have seen. Mis-behaving box made available by: John Cagle <john.cagle@hp.com>
2004-04-06 18:28:15 +00:00
CSR_WRITE_4(sc, BGE_MARB_MODE, BGE_MARBMODE_ENABLE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Prevent PXE restart: write a magic number to the
* general communications memory at 0xB50.
*/
bge_writemem_ind(sc, BGE_SOFTWARE_GENCOMM, BGE_MAGIC_NUMBER);
/*
* Poll the value location we just wrote until
* we see the 1's complement of the magic number.
* This indicates that the firmware initialization
* is complete.
*/
for (i = 0; i < BGE_TIMEOUT; i++) {
val = bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM);
if (val == ~BGE_MAGIC_NUMBER)
break;
DELAY(10);
}
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "firmware handshake timed out\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
/*
* XXX Wait for the value of the PCISTATE register to
* return to its original pre-reset state. This is a
* fairly good indicator of reset completion. If we don't
* wait for the reset to fully complete, trying to read
* from the device's non-PCI registers may yield garbage
* results.
*/
for (i = 0; i < BGE_TIMEOUT; i++) {
if (pci_read_config(dev, BGE_PCI_PCISTATE, 4) == pcistate)
break;
DELAY(10);
}
/* Fix up byte swapping */
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
CSR_WRITE_4(sc, BGE_MODE_CTL, BGE_DMA_SWAP_OPTIONS|
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_MODECTL_BYTESWAP_DATA);
CSR_WRITE_4(sc, BGE_MAC_MODE, 0);
Add some special case code to fix a problem with the BCM5704 in TBI (fiber) mode. The 5704 apparently has some s00p3r s33kr1t registers for setting the advertisement of pause frame ability (i.e flow control) when in autoneg mode. If we don't set these registers correctly, we may not be able to negotiate a proper link with some switches. (Symptom is that the NIC reports the link as up (PCS synched) but no traffic can be exchanged.) PR: kern/67598
2004-06-09 16:01:59 +00:00
/*
* The 5704 in TBI mode apparently needs some special
* adjustment to insure the SERDES drive level is set
* to 1.2V.
*/
if (sc->bge_asicrev == BGE_ASICREV_BCM5704 && sc->bge_tbi) {
uint32_t serdescfg;
serdescfg = CSR_READ_4(sc, BGE_SERDES_CFG);
serdescfg = (serdescfg & ~0xFFF) | 0x880;
CSR_WRITE_4(sc, BGE_SERDES_CFG, serdescfg);
}
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
/* XXX: Broadcom Linux driver. */
if (sc->bge_pcie && sc->bge_chipid != BGE_CHIPID_BCM5750_A0) {
uint32_t v;
v = CSR_READ_4(sc, 0x7c00);
CSR_WRITE_4(sc, 0x7c00, v | (1<<25));
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
DELAY(10000);
return;
}
/*
* Frame reception handling. This is called if there's a frame
* on the receive return list.
*
* Note: we have to be able to handle two possibilities here:
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
* 1) the frame is from the jumbo receive ring
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* 2) the frame is from the standard receive ring
*/
static void
bge_rxeof(sc)
struct bge_softc *sc;
{
struct ifnet *ifp;
int stdcnt = 0, jumbocnt = 0;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_return_ring_tag,
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
sc->bge_cdata.bge_rx_return_ring_map, BUS_DMASYNC_POSTREAD);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_POSTREAD);
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map,
BUS_DMASYNC_POSTREAD);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
while(sc->bge_rx_saved_considx !=
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
struct bge_rx_bd *cur_rx;
u_int32_t rxidx;
struct ether_header *eh;
struct mbuf *m = NULL;
u_int16_t vlan_tag = 0;
int have_tag = 0;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING) {
if (sc->rxcycles <= 0)
break;
sc->rxcycles--;
}
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
cur_rx =
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
&sc->bge_ldata.bge_rx_return_ring[sc->bge_rx_saved_considx];
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
rxidx = cur_rx->bge_idx;
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_INC(sc->bge_rx_saved_considx, sc->bge_return_ring_cnt);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (cur_rx->bge_flags & BGE_RXBDFLAG_VLAN_TAG) {
have_tag = 1;
vlan_tag = cur_rx->bge_vlan_tag;
}
if (cur_rx->bge_flags & BGE_RXBDFLAG_JUMBO_RING) {
BGE_INC(sc->bge_jumbo, BGE_JUMBO_RX_RING_CNT);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[rxidx],
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[rxidx]);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m = sc->bge_cdata.bge_rx_jumbo_chain[rxidx];
sc->bge_cdata.bge_rx_jumbo_chain[rxidx] = NULL;
jumbocnt++;
if (cur_rx->bge_flags & BGE_RXBDFLAG_ERROR) {
ifp->if_ierrors++;
bge_newbuf_jumbo(sc, sc->bge_jumbo, m);
continue;
}
if (bge_newbuf_jumbo(sc,
sc->bge_jumbo, NULL) == ENOBUFS) {
ifp->if_ierrors++;
bge_newbuf_jumbo(sc, sc->bge_jumbo, m);
continue;
}
} else {
BGE_INC(sc->bge_std, BGE_STD_RX_RING_CNT);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[rxidx],
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[rxidx]);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m = sc->bge_cdata.bge_rx_std_chain[rxidx];
sc->bge_cdata.bge_rx_std_chain[rxidx] = NULL;
stdcnt++;
if (cur_rx->bge_flags & BGE_RXBDFLAG_ERROR) {
ifp->if_ierrors++;
bge_newbuf_std(sc, sc->bge_std, m);
continue;
}
if (bge_newbuf_std(sc, sc->bge_std,
NULL) == ENOBUFS) {
ifp->if_ierrors++;
bge_newbuf_std(sc, sc->bge_std, m);
continue;
}
}
ifp->if_ipackets++;
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
#ifndef __NO_STRICT_ALIGNMENT
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
/*
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
* For architectures with strict alignment we must make sure
* the payload is aligned.
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
*/
if (sc->bge_rx_alignment_bug) {
bcopy(m->m_data, m->m_data + ETHER_ALIGN,
cur_rx->bge_len);
m->m_data += ETHER_ALIGN;
}
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
eh = mtod(m, struct ether_header *);
By not setting No_CRC in the Mode Control Register, we must also reduce the size of the packet by 4 bytes to remove the ethernet crc. Submitted by: jdp
2003-09-23 19:54:32 +00:00
m->m_pkthdr.len = m->m_len = cur_rx->bge_len - ETHER_CRC_LEN;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m->m_pkthdr.rcvif = ifp;
Bring the advertised interface capabilities into line with the reality (in particular, bge(4) hasn't supported rxcsum since if_bge.c#1.5) Clean up some aspects of capabilities usage, i.e. stop using if_hwassist to see whether we are doing offload now because if_hwassist is for TCP/IP layer and it is subordinate to if_capenable. Thanks to: Aled Morris for donating a nice bge(4) NIC to me Reviewed by: -net, -hackers (silence)
2004-07-24 13:45:38 +00:00
if (ifp->if_capenable & IFCAP_RXCSUM) {
Check for BGE_RXBDFLAG_IP_CSUM flag before marking mbuf as with valid checksum. This fixes operation of PPPoE and other non-IP Ethernet protocols.
2006-01-21 09:54:32 +00:00
if (cur_rx->bge_flags & BGE_RXBDFLAG_IP_CSUM) {
m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
if ((cur_rx->bge_ip_csum ^ 0xffff) == 0)
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
}
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
if (cur_rx->bge_flags & BGE_RXBDFLAG_TCP_UDP_CSUM &&
m->m_pkthdr.len >= ETHER_MIN_NOPAD) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m->m_pkthdr.csum_data =
cur_rx->bge_tcp_udp_csum;
Set the 'no pseudo header checksum' option for RX checksums, otherwise we may botch UDP checksums on receive, which will break NFS.
2001-09-28 18:56:57 +00:00
m->m_pkthdr.csum_flags |= CSUM_DATA_VALID;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
}
/*
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
* If we received a packet with a vlan tag,
* attach that information to the packet.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
- Fix VLAN_INPUT_TAG() macro, so that it doesn't touch mtag in case if memory allocation failed. - Remove fourth argument from VLAN_INPUT_TAG(), that was used incorrectly in almost all drivers. Indicate failure with mbuf value of NULL. In collaboration with: yongari, ru, sam
2005-12-18 18:24:27 +00:00
if (have_tag) {
VLAN_INPUT_TAG(ifp, m, vlan_tag);
if (m == NULL)
continue;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_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);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (stdcnt > 0)
bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_PREWRITE);
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (jumbocnt > 0)
bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map,
BUS_DMASYNC_PREWRITE);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MBX_RX_CONS0_LO, sc->bge_rx_saved_considx);
if (stdcnt)
CSR_WRITE_4(sc, BGE_MBX_RX_STD_PROD_LO, sc->bge_std);
if (jumbocnt)
CSR_WRITE_4(sc, BGE_MBX_RX_JUMBO_PROD_LO, sc->bge_jumbo);
return;
}
static void
bge_txeof(sc)
struct bge_softc *sc;
{
struct bge_tx_bd *cur_tx = NULL;
struct ifnet *ifp;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_cdata.bge_tx_ring_map,
BUS_DMASYNC_POSTREAD|BUS_DMASYNC_POSTWRITE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Go through our tx ring and free mbufs for those
* frames that have been sent.
*/
while (sc->bge_tx_saved_considx !=
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
u_int32_t idx = 0;
idx = sc->bge_tx_saved_considx;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
cur_tx = &sc->bge_ldata.bge_tx_ring[idx];
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (cur_tx->bge_flags & BGE_TXBDFLAG_END)
ifp->if_opackets++;
if (sc->bge_cdata.bge_tx_chain[idx] != NULL) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[idx],
BUS_DMASYNC_POSTWRITE);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[idx]);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
m_freem(sc->bge_cdata.bge_tx_chain[idx]);
sc->bge_cdata.bge_tx_chain[idx] = NULL;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
sc->bge_txcnt--;
BGE_INC(sc->bge_tx_saved_considx, BGE_TX_RING_CNT);
ifp->if_timer = 0;
}
if (cur_tx != NULL)
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
static void
bge_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
{
struct bge_softc *sc = ifp->if_softc;
BGE_LOCK(sc);
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
bge_poll_locked(ifp, cmd, count);
BGE_UNLOCK(sc);
}
static void
bge_poll_locked(struct ifnet *ifp, enum poll_cmd cmd, int count)
{
struct bge_softc *sc = ifp->if_softc;
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
uint32_t statusword;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
BGE_LOCK_ASSERT(sc);
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_POSTREAD);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
statusword = atomic_readandclear_32(&sc->bge_ldata.bge_status_block->bge_status);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_PREREAD);
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
/* Note link event. It will be processed by POLL_AND_CHECK_STATUS cmd */
if (statusword & BGE_STATFLAG_LINKSTATE_CHANGED)
sc->bge_link_evt++;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
if (cmd == POLL_AND_CHECK_STATUS)
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if ((sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
sc->bge_chipid != BGE_CHIPID_BCM5700_B1) ||
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
sc->bge_link_evt || sc->bge_tbi)
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
bge_link_upd(sc);
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
sc->rxcycles = count;
bge_rxeof(sc);
bge_txeof(sc);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
bge_start_locked(ifp);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
}
#endif /* DEVICE_POLLING */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static void
bge_intr(xsc)
void *xsc;
{
struct bge_softc *sc;
struct ifnet *ifp;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
uint32_t statusword;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc = xsc;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
ifp = sc->bge_ifp;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING) {
BGE_UNLOCK(sc);
return;
}
#endif
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
sc->bge_cdata.bge_status_map, BUS_DMASYNC_POSTREAD);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
- Make this work with PAE. - atomically load and clear the status block so we dont miss an update. Submitted by: jdp Approved by: re (scottl)
2003-05-21 07:00:49 +00:00
statusword =
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
atomic_readandclear_32(&sc->bge_ldata.bge_status_block->bge_status);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_PREREAD);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#ifdef notdef
/* Avoid this for now -- checking this register is expensive. */
/* Make sure this is really our interrupt. */
if (!(CSR_READ_4(sc, BGE_MISC_LOCAL_CTL) & BGE_MLC_INTR_STATE))
return;
#endif
/* Ack interrupt and stop others from occuring. */
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 1);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if ((sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
sc->bge_chipid != BGE_CHIPID_BCM5700_B1) ||
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
statusword & BGE_STATFLAG_LINKSTATE_CHANGED || sc->bge_link_evt)
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
bge_link_upd(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Check RX return ring producer/consumer */
bge_rxeof(sc);
/* Check TX ring producer/consumer */
bge_txeof(sc);
}
/* Re-enable interrupts. */
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 0);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
bge_start_locked(ifp);
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
static void
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
bge_tick_locked(sc)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
struct bge_softc *sc;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
{
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
struct mii_data *mii = NULL;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5705 ||
sc->bge_asicrev == BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
bge_stats_update_regs(sc);
else
bge_stats_update(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if (!sc->bge_tbi) {
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
mii = device_get_softc(sc->bge_miibus);
mii_tick(mii);
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
} else {
/*
* Since in TBI mode auto-polling can't be used we should poll
* link status manually. Here we register pending link event
* and trigger interrupt.
*/
#ifdef DEVICE_POLLING
/* In polling mode we poll link state in bge_poll_locked() */
if (!(sc->bge_ifp->if_capenable & IFCAP_POLLING))
#endif
{
sc->bge_link_evt++;
BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_INTR_SET);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
callout_reset(&sc->bge_stat_ch, hz, bge_tick, sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
static void
bge_tick(xsc)
void *xsc;
{
struct bge_softc *sc;
sc = xsc;
BGE_LOCK(sc);
bge_tick_locked(sc);
BGE_UNLOCK(sc);
}
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
static void
bge_stats_update_regs(sc)
struct bge_softc *sc;
{
struct ifnet *ifp;
struct bge_mac_stats_regs stats;
u_int32_t *s;
- Count packets discarded by RX/TX MAC (cause of FIFO overflow, etc) as input/output interface errors. - Keep values of rx/tx discards & tx collisions inside struct bge_softc. So we can keep statistic across ifconfig down/up runs (cause bringing bge up will reset chip). Approved by: glebius (mentor) MFC after: 1 week
2006-01-17 23:01:58 +00:00
u_long cnt; /* current register value */
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
int i;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
s = (u_int32_t *)&stats;
for (i = 0; i < sizeof(struct bge_mac_stats_regs); i += 4) {
*s = CSR_READ_4(sc, BGE_RX_STATS + i);
s++;
}
- Count packets discarded by RX/TX MAC (cause of FIFO overflow, etc) as input/output interface errors. - Keep values of rx/tx discards & tx collisions inside struct bge_softc. So we can keep statistic across ifconfig down/up runs (cause bringing bge up will reset chip). Approved by: glebius (mentor) MFC after: 1 week
2006-01-17 23:01:58 +00:00
cnt = stats.dot3StatsSingleCollisionFrames +
stats.dot3StatsMultipleCollisionFrames +
stats.dot3StatsExcessiveCollisions +
stats.dot3StatsLateCollisions;
ifp->if_collisions += cnt >= sc->bge_tx_collisions ?
cnt - sc->bge_tx_collisions : cnt;
sc->bge_tx_collisions = cnt;
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static void
bge_stats_update(sc)
struct bge_softc *sc;
{
struct ifnet *ifp;
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
bus_size_t stats;
- Count packets discarded by RX/TX MAC (cause of FIFO overflow, etc) as input/output interface errors. - Keep values of rx/tx discards & tx collisions inside struct bge_softc. So we can keep statistic across ifconfig down/up runs (cause bringing bge up will reset chip). Approved by: glebius (mentor) MFC after: 1 week
2006-01-17 23:01:58 +00:00
u_long cnt; /* current register value */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
stats = BGE_MEMWIN_START + BGE_STATS_BLOCK;
#define READ_STAT(sc, stats, stat) \
CSR_READ_4(sc, stats + offsetof(struct bge_stats, stat))
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- Count packets discarded by RX/TX MAC (cause of FIFO overflow, etc) as input/output interface errors. - Keep values of rx/tx discards & tx collisions inside struct bge_softc. So we can keep statistic across ifconfig down/up runs (cause bringing bge up will reset chip). Approved by: glebius (mentor) MFC after: 1 week
2006-01-17 23:01:58 +00:00
cnt = READ_STAT(sc, stats,
txstats.dot3StatsSingleCollisionFrames.bge_addr_lo);
cnt += READ_STAT(sc, stats,
txstats.dot3StatsMultipleCollisionFrames.bge_addr_lo);
cnt += READ_STAT(sc, stats,
txstats.dot3StatsExcessiveCollisions.bge_addr_lo);
cnt += READ_STAT(sc, stats,
txstats.dot3StatsLateCollisions.bge_addr_lo);
ifp->if_collisions += cnt >= sc->bge_tx_collisions ?
cnt - sc->bge_tx_collisions : cnt;
sc->bge_tx_collisions = cnt;
cnt = READ_STAT(sc, stats, ifInDiscards.bge_addr_lo);
ifp->if_ierrors += cnt >= sc->bge_rx_discards ?
cnt - sc->bge_rx_discards : cnt;
sc->bge_rx_discards = cnt;
cnt = READ_STAT(sc, stats, txstats.ifOutDiscards.bge_addr_lo);
ifp->if_oerrors += cnt >= sc->bge_tx_discards ?
cnt - sc->bge_tx_discards : cnt;
sc->bge_tx_discards = cnt;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
#undef READ_STAT
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
/*
* Pad outbound frame to ETHER_MIN_NOPAD for an unusual reason.
* The bge hardware will pad out Tx runts to ETHER_MIN_NOPAD,
* but when such padded frames employ the bge IP/TCP checksum offload,
* the hardware checksum assist gives incorrect results (possibly
* from incorporating its own padding into the UDP/TCP checksum; who knows).
* If we pad such runts with zeros, the onboard checksum comes out correct.
*/
static __inline int
bge_cksum_pad(struct mbuf *m)
{
int padlen = ETHER_MIN_NOPAD - m->m_pkthdr.len;
struct mbuf *last;
/* If there's only the packet-header and we can pad there, use it. */
if (m->m_pkthdr.len == m->m_len && M_WRITABLE(m) &&
M_TRAILINGSPACE(m) >= padlen) {
last = m;
} else {
/*
* Walk packet chain to find last mbuf. We will either
* pad there, or append a new mbuf and pad it.
*/
for (last = m; last->m_next != NULL; last = last->m_next);
if (!(M_WRITABLE(last) && M_TRAILINGSPACE(last) >= padlen)) {
/* Allocate new empty mbuf, pad it. Compact later. */
struct mbuf *n;
MGET(n, M_DONTWAIT, MT_DATA);
if (n == NULL)
return (ENOBUFS);
n->m_len = 0;
last->m_next = n;
last = n;
}
}
/* Now zero the pad area, to avoid the bge cksum-assist bug. */
memset(mtod(last, caddr_t) + last->m_len, 0, padlen);
last->m_len += padlen;
m->m_pkthdr.len += padlen;
return (0);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Encapsulate an mbuf chain in the tx ring by coupling the mbuf data
* pointers to descriptors.
*/
static int
bge_encap(sc, m_head, txidx)
struct bge_softc *sc;
struct mbuf *m_head;
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
uint32_t *txidx;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
bus_dma_segment_t segs[BGE_NSEG_NEW];
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_t map;
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
struct bge_tx_bd *d = NULL;
struct m_tag *mtag;
uint32_t idx = *txidx;
uint16_t csum_flags = 0;
int nsegs, i, error;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (m_head->m_pkthdr.csum_flags) {
if (m_head->m_pkthdr.csum_flags & CSUM_IP)
csum_flags |= BGE_TXBDFLAG_IP_CSUM;
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
if (m_head->m_pkthdr.csum_flags & (CSUM_TCP | CSUM_UDP)) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
csum_flags |= BGE_TXBDFLAG_TCP_UDP_CSUM;
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
if (m_head->m_pkthdr.len < ETHER_MIN_NOPAD &&
bge_cksum_pad(m_head) != 0)
return (ENOBUFS);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (m_head->m_flags & M_LASTFRAG)
csum_flags |= BGE_TXBDFLAG_IP_FRAG_END;
else if (m_head->m_flags & M_FRAG)
csum_flags |= BGE_TXBDFLAG_IP_FRAG;
}
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
mtag = VLAN_OUTPUT_TAG(sc->bge_ifp, m_head);
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
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
map = sc->bge_cdata.bge_tx_dmamap[idx];
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_mtag, map,
m_head, segs, &nsegs, BUS_DMA_NOWAIT);
if (error) {
if (error == EFBIG) {
struct mbuf *m0;
m0 = m_defrag(m_head, M_DONTWAIT);
if (m0 == NULL)
return (ENOBUFS);
m_head = m0;
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_mtag,
map, m_head, segs, &nsegs, BUS_DMA_NOWAIT);
}
if (error)
return (error);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
* Sanity check: avoid coming within 16 descriptors
* of the end of the ring.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
if (nsegs > (BGE_TX_RING_CNT - sc->bge_txcnt - 16)) {
bus_dmamap_unload(sc->bge_cdata.bge_mtag, map);
return (ENOBUFS);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag, map, BUS_DMASYNC_PREWRITE);
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
for (i = 0; ; i++) {
d = &sc->bge_ldata.bge_tx_ring[idx];
d->bge_addr.bge_addr_lo = BGE_ADDR_LO(segs[i].ds_addr);
d->bge_addr.bge_addr_hi = BGE_ADDR_HI(segs[i].ds_addr);
d->bge_len = segs[i].ds_len;
d->bge_flags = csum_flags;
if (i == nsegs - 1)
break;
BGE_INC(idx, BGE_TX_RING_CNT);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
/* Mark the last segment as end of packet... */
d->bge_flags |= BGE_TXBDFLAG_END;
/* ... and put VLAN tag into first segment. */
d = &sc->bge_ldata.bge_tx_ring[*txidx];
if (mtag != NULL) {
d->bge_flags |= BGE_TXBDFLAG_VLAN_TAG;
d->bge_vlan_tag = VLAN_TAG_VALUE(mtag);
} else
d->bge_vlan_tag = 0;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/*
* Insure that the map for this transmission
* is placed at the array index of the last descriptor
* in this chain.
*/
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
sc->bge_cdata.bge_tx_dmamap[*txidx] = sc->bge_cdata.bge_tx_dmamap[idx];
sc->bge_cdata.bge_tx_dmamap[idx] = map;
sc->bge_cdata.bge_tx_chain[idx] = m_head;
sc->bge_txcnt += nsegs;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
BGE_INC(idx, BGE_TX_RING_CNT);
*txidx = idx;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* Main transmit routine. To avoid having to do mbuf copies, we put pointers
* to the mbuf data regions directly in the transmit descriptors.
*/
static void
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
bge_start_locked(ifp)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
struct ifnet *ifp;
{
struct bge_softc *sc;
struct mbuf *m_head = NULL;
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
uint32_t prodidx;
With ALTQ, it is possible that although the queue was not empty when we entered the interface start function, no packets were actually dequeued. Therefore, keep a count of how many packets we really added onto the tx chain, and initiate a transmit only if the count is non-zero.
2004-10-30 22:59:30 +00:00
int count = 0;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc = ifp->if_softc;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
if (!sc->bge_link || IFQ_DRV_IS_EMPTY(&ifp->if_snd))
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
prodidx = sc->bge_tx_prodidx;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
while(sc->bge_cdata.bge_tx_chain[prodidx] == NULL) {
Add altq(4) support.
2004-10-30 21:21:10 +00:00
IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (m_head == NULL)
break;
/*
Bring the advertised interface capabilities into line with the reality (in particular, bge(4) hasn't supported rxcsum since if_bge.c#1.5) Clean up some aspects of capabilities usage, i.e. stop using if_hwassist to see whether we are doing offload now because if_hwassist is for TCP/IP layer and it is subordinate to if_capenable. Thanks to: Aled Morris for donating a nice bge(4) NIC to me Reviewed by: -net, -hackers (silence)
2004-07-24 13:45:38 +00:00
* XXX
* The code inside the if() block is never reached since we
* must mark CSUM_IP_FRAGS in our if_hwassist to start getting
* requests to checksum TCP/UDP in a fragmented packet.
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
*
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* XXX
* safety overkill. If this is a fragmented packet chain
* with delayed TCP/UDP checksums, then only encapsulate
* it if we have enough descriptors to handle the entire
* chain at once.
* (paranoia -- may not actually be needed)
*/
if (m_head->m_flags & M_FIRSTFRAG &&
m_head->m_pkthdr.csum_flags & (CSUM_DELAY_DATA)) {
if ((BGE_TX_RING_CNT - sc->bge_txcnt) <
m_head->m_pkthdr.csum_data + 16) {
Add altq(4) support.
2004-10-30 21:21:10 +00:00
IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
break;
}
}
/*
* Pack the data into the transmit ring. If we
* don't have room, set the OACTIVE flag and wait
* for the NIC to drain the ring.
*/
if (bge_encap(sc, m_head, &prodidx)) {
Add altq(4) support.
2004-10-30 21:21:10 +00:00
IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
break;
}
With ALTQ, it is possible that although the queue was not empty when we entered the interface start function, no packets were actually dequeued. Therefore, keep a count of how many packets we really added onto the tx chain, and initiate a transmit only if the count is non-zero.
2004-10-30 22:59:30 +00:00
++count;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* If there's a BPF listener, bounce a copy of this frame
* to him.
*/
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, m_head);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
With ALTQ, it is possible that although the queue was not empty when we entered the interface start function, no packets were actually dequeued. Therefore, keep a count of how many packets we really added onto the tx chain, and initiate a transmit only if the count is non-zero.
2004-10-30 22:59:30 +00:00
if (count == 0) {
/* no packets were dequeued */
return;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Transmit */
CSR_WRITE_4(sc, BGE_MBX_TX_HOST_PROD0_LO, prodidx);
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
/* 5700 b2 errata */
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
CSR_WRITE_4(sc, BGE_MBX_TX_HOST_PROD0_LO, prodidx);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
sc->bge_tx_prodidx = prodidx;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Set a timeout in case the chip goes out to lunch.
*/
ifp->if_timer = 5;
return;
}
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
/*
* Main transmit routine. To avoid having to do mbuf copies, we put pointers
* to the mbuf data regions directly in the transmit descriptors.
*/
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static void
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
bge_start(ifp)
struct ifnet *ifp;
{
struct bge_softc *sc;
sc = ifp->if_softc;
BGE_LOCK(sc);
bge_start_locked(ifp);
BGE_UNLOCK(sc);
}
static void
bge_init_locked(sc)
struct bge_softc *sc;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
struct ifnet *ifp;
u_int16_t *m;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
/* Cancel pending I/O and flush buffers. */
bge_stop(sc);
bge_reset(sc);
bge_chipinit(sc);
/*
* Init the various state machines, ring
* control blocks and firmware.
*/
if (bge_blockinit(sc)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "initialization failure\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Specify MTU. */
CSR_WRITE_4(sc, BGE_RX_MTU, ifp->if_mtu +
Set the BGE_RX_MTU register correctly so that we can receive slightly larger than normal frames, to account for the case where a bge(4) NIC is used with VLANs. Since we set the IFCAP_VLAN_MTU flag, we must allow reception of frames up to 1522 bytes in size rather than 1518. Note that it is possible to work around this bug by doing: # ifconfig bge0 mtu 1504 prior to configuring any VLAN interfaces.
2003-08-04 05:50:53 +00:00
ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Load our MAC address. */
- Store pointer to the link-level address right in "struct ifnet" rather than in ifindex_table[]; all (except one) accesses are through ifp anyway. IF_LLADDR() works faster, and all (except one) ifaddr_byindex() users were converted to use ifp->if_addr. - Stop storing a (pointer to) Ethernet address in "struct arpcom", and drop the IFP2ENADDR() macro; all users have been converted to use IF_LLADDR() instead.
2005-11-11 16:04:59 +00:00
m = (u_int16_t *)IF_LLADDR(sc->bge_ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MAC_ADDR1_LO, htons(m[0]));
CSR_WRITE_4(sc, BGE_MAC_ADDR1_HI, (htons(m[1]) << 16) | htons(m[2]));
/* Enable or disable promiscuous mode as needed. */
if (ifp->if_flags & IFF_PROMISC) {
BGE_SETBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_PROMISC);
} else {
BGE_CLRBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_PROMISC);
}
/* Program multicast filter. */
bge_setmulti(sc);
/* Init RX ring. */
bge_init_rx_ring_std(sc);
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
/*
* Workaround for a bug in 5705 ASIC rev A0. Poll the NIC's
* memory to insure that the chip has in fact read the first
* entry of the ring.
*/
if (sc->bge_chipid == BGE_CHIPID_BCM5705_A0) {
u_int32_t v, i;
for (i = 0; i < 10; i++) {
DELAY(20);
v = bge_readmem_ind(sc, BGE_STD_RX_RINGS + 8);
if (v == (MCLBYTES - ETHER_ALIGN))
break;
}
if (i == 10)
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf (sc->bge_dev,
"5705 A0 chip failed to load RX ring\n");
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Init jumbo RX ring. */
if (ifp->if_mtu > (ETHERMTU + ETHER_HDR_LEN + ETHER_CRC_LEN))
bge_init_rx_ring_jumbo(sc);
/* Init our RX return ring index */
sc->bge_rx_saved_considx = 0;
/* Init TX ring. */
bge_init_tx_ring(sc);
/* Turn on transmitter */
BGE_SETBIT(sc, BGE_TX_MODE, BGE_TXMODE_ENABLE);
/* Turn on receiver */
BGE_SETBIT(sc, BGE_RX_MODE, BGE_RXMODE_ENABLE);
/* Tell firmware we're alive. */
BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
/* Disable interrupts if we are polling. */
if (ifp->if_capenable & IFCAP_POLLING) {
BGE_SETBIT(sc, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 1);
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 1);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 1);
} else
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Enable host interrupts. */
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
{
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_SETBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_CLEAR_INTA);
BGE_CLRBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 0);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_ifmedia_upd(ifp);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
callout_reset(&sc->bge_stat_ch, hz, bge_tick, sc);
}
static void
bge_init(xsc)
void *xsc;
{
struct bge_softc *sc = xsc;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
bge_init_locked(sc);
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
/*
* Set media options.
*/
static int
bge_ifmedia_upd(ifp)
struct ifnet *ifp;
{
struct bge_softc *sc;
struct mii_data *mii;
struct ifmedia *ifm;
sc = ifp->if_softc;
ifm = &sc->bge_ifmedia;
/* If this is a 1000baseX NIC, enable the TBI port. */
if (sc->bge_tbi) {
if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
return(EINVAL);
switch(IFM_SUBTYPE(ifm->ifm_media)) {
case IFM_AUTO:
Backout the change I made before 5.4-R since I wasn't aware that it was only a problem with one particular switch module. Create a kernel option BGE_FAKE_AUTONEG that restores the 5.4 behavior, which should make the DNLK switch module work. IBM/Intel blades with Intel or AD switch modules should work without patching or kernel options with this commit. Hardware for testing provided by several folks, including Danny Braniss <danny@cs.huji.ac.il>, Achim Patzner <ap@bnc.net>, and OffMyServer. Approved by: re
2005-06-24 21:43:47 +00:00
#ifndef BGE_FAKE_AUTONEG
/*
* The BCM5704 ASIC appears to have a special
* mechanism for programming the autoneg
* advertisement registers in TBI mode.
*/
if (sc->bge_asicrev == BGE_ASICREV_BCM5704) {
uint32_t sgdig;
CSR_WRITE_4(sc, BGE_TX_TBI_AUTONEG, 0);
sgdig = CSR_READ_4(sc, BGE_SGDIG_CFG);
sgdig |= BGE_SGDIGCFG_AUTO|
BGE_SGDIGCFG_PAUSE_CAP|
BGE_SGDIGCFG_ASYM_PAUSE;
CSR_WRITE_4(sc, BGE_SGDIG_CFG,
sgdig|BGE_SGDIGCFG_SEND);
DELAY(5);
CSR_WRITE_4(sc, BGE_SGDIG_CFG, sgdig);
}
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
break;
case IFM_1000_SX:
if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) {
BGE_CLRBIT(sc, BGE_MAC_MODE,
BGE_MACMODE_HALF_DUPLEX);
} else {
BGE_SETBIT(sc, BGE_MAC_MODE,
BGE_MACMODE_HALF_DUPLEX);
}
break;
default:
return(EINVAL);
}
return(0);
}
mii = device_get_softc(sc->bge_miibus);
if (mii->mii_instance) {
struct mii_softc *miisc;
for (miisc = LIST_FIRST(&mii->mii_phys); miisc != NULL;
miisc = LIST_NEXT(miisc, mii_list))
mii_phy_reset(miisc);
}
mii_mediachg(mii);
return(0);
}
/*
* Report current media status.
*/
static void
bge_ifmedia_sts(ifp, ifmr)
struct ifnet *ifp;
struct ifmediareq *ifmr;
{
struct bge_softc *sc;
struct mii_data *mii;
sc = ifp->if_softc;
if (sc->bge_tbi) {
ifmr->ifm_status = IFM_AVALID;
ifmr->ifm_active = IFM_ETHER;
if (CSR_READ_4(sc, BGE_MAC_STS) &
BGE_MACSTAT_TBI_PCS_SYNCHED)
ifmr->ifm_status |= IFM_ACTIVE;
ifmr->ifm_active |= IFM_1000_SX;
if (CSR_READ_4(sc, BGE_MAC_MODE) & BGE_MACMODE_HALF_DUPLEX)
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
ifmr->ifm_active |= IFM_HDX;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
else
ifmr->ifm_active |= IFM_FDX;
return;
}
mii = device_get_softc(sc->bge_miibus);
mii_pollstat(mii);
ifmr->ifm_active = mii->mii_media_active;
ifmr->ifm_status = mii->mii_media_status;
return;
}
static int
bge_ioctl(ifp, command, data)
struct ifnet *ifp;
u_long command;
caddr_t data;
{
struct bge_softc *sc = ifp->if_softc;
struct ifreq *ifr = (struct ifreq *) data;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
int mask, error = 0;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
struct mii_data *mii;
switch(command) {
case SIOCSIFMTU:
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
/* Disallow jumbo frames on 5705. */
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
if (((sc->bge_asicrev == BGE_ASICREV_BCM5705 ||
sc->bge_asicrev == BGE_ASICREV_BCM5750) &&
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
ifr->ifr_mtu > ETHERMTU) || ifr->ifr_mtu > BGE_JUMBO_MTU)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = EINVAL;
else {
ifp->if_mtu = ifr->ifr_mtu;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_init(sc);
}
break;
case SIOCSIFFLAGS:
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (ifp->if_flags & IFF_UP) {
/*
* If only the state of the PROMISC flag changed,
* then just use the 'set promisc mode' command
* instead of reinitializing the entire NIC. Doing
* a full re-init means reloading the firmware and
* waiting for it to start up, which may take a
Setting IFF_ALLMULTI on a running interface didn't call bge_setmulti() as intended, resulting in a non-working multicast routing. Fix it. MFC after: 3 days
2006-02-01 10:11:24 +00:00
* second or two. Similarly for ALLMULTI.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_flags & IFF_PROMISC &&
!(sc->bge_if_flags & IFF_PROMISC)) {
BGE_SETBIT(sc, BGE_RX_MODE,
BGE_RXMODE_RX_PROMISC);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
} else if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
!(ifp->if_flags & IFF_PROMISC) &&
sc->bge_if_flags & IFF_PROMISC) {
BGE_CLRBIT(sc, BGE_RX_MODE,
BGE_RXMODE_RX_PROMISC);
Setting IFF_ALLMULTI on a running interface didn't call bge_setmulti() as intended, resulting in a non-working multicast routing. Fix it. MFC after: 3 days
2006-02-01 10:11:24 +00:00
} else if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
(ifp->if_flags ^ sc->bge_if_flags) & IFF_ALLMULTI) {
bge_setmulti(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
} else
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
bge_init_locked(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
} else {
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_stop(sc);
}
}
sc->bge_if_flags = ifp->if_flags;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = 0;
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_setmulti(sc);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = 0;
}
break;
case SIOCSIFMEDIA:
case SIOCGIFMEDIA:
if (sc->bge_tbi) {
error = ifmedia_ioctl(ifp, ifr,
&sc->bge_ifmedia, command);
} else {
mii = device_get_softc(sc->bge_miibus);
error = ifmedia_ioctl(ifp, ifr,
&mii->mii_media, command);
}
break;
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
case SIOCSIFCAP:
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
mask = ifr->ifr_reqcap ^ ifp->if_capenable;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
if (mask & IFCAP_POLLING) {
if (ifr->ifr_reqcap & IFCAP_POLLING) {
error = ether_poll_register(bge_poll, ifp);
if (error)
return(error);
BGE_LOCK(sc);
BGE_SETBIT(sc, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 1);
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 1);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 1);
ifp->if_capenable |= IFCAP_POLLING;
BGE_UNLOCK(sc);
} else {
error = ether_poll_deregister(ifp);
/* Enable interrupt even in error case */
BGE_LOCK(sc);
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 0);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 0);
BGE_CLRBIT(sc, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 0);
ifp->if_capenable &= ~IFCAP_POLLING;
BGE_UNLOCK(sc);
}
}
#endif
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
if (mask & IFCAP_HWCSUM) {
ifp->if_capenable ^= IFCAP_HWCSUM;
if (IFCAP_HWCSUM & ifp->if_capenable &&
IFCAP_HWCSUM & ifp->if_capabilities)
Bring the advertised interface capabilities into line with the reality (in particular, bge(4) hasn't supported rxcsum since if_bge.c#1.5) Clean up some aspects of capabilities usage, i.e. stop using if_hwassist to see whether we are doing offload now because if_hwassist is for TCP/IP layer and it is subordinate to if_capenable. Thanks to: Aled Morris for donating a nice bge(4) NIC to me Reviewed by: -net, -hackers (silence)
2004-07-24 13:45:38 +00:00
ifp->if_hwassist = BGE_CSUM_FEATURES;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
else
Bring the advertised interface capabilities into line with the reality (in particular, bge(4) hasn't supported rxcsum since if_bge.c#1.5) Clean up some aspects of capabilities usage, i.e. stop using if_hwassist to see whether we are doing offload now because if_hwassist is for TCP/IP layer and it is subordinate to if_capenable. Thanks to: Aled Morris for donating a nice bge(4) NIC to me Reviewed by: -net, -hackers (silence)
2004-07-24 13:45:38 +00:00
ifp->if_hwassist = 0;
This driver can do hardware VLAN tagging + checksum offloading. In collaboration with: Mihail Balikov <mihail.balikov interbgc.com>
2006-01-30 13:45:55 +00:00
VLAN_CAPABILITIES(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
break;
default:
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
error = ether_ioctl(ifp, command, data);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
break;
}
return(error);
}
static void
bge_watchdog(ifp)
struct ifnet *ifp;
{
struct bge_softc *sc;
sc = ifp->if_softc;
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
if_printf(ifp, "watchdog timeout -- resetting\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_init(sc);
ifp->if_oerrors++;
return;
}
/*
* Stop the adapter and free any mbufs allocated to the
* RX and TX lists.
*/
static void
bge_stop(sc)
struct bge_softc *sc;
{
struct ifnet *ifp;
struct ifmedia_entry *ifm;
struct mii_data *mii = NULL;
int mtmp, itmp;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (!sc->bge_tbi)
mii = device_get_softc(sc->bge_miibus);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
callout_stop(&sc->bge_stat_ch);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Disable all of the receiver blocks
*/
BGE_CLRBIT(sc, BGE_RX_MODE, BGE_RXMODE_ENABLE);
BGE_CLRBIT(sc, BGE_RBDI_MODE, BGE_RBDIMODE_ENABLE);
BGE_CLRBIT(sc, BGE_RXLP_MODE, BGE_RXLPMODE_ENABLE);
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_CLRBIT(sc, BGE_RXLS_MODE, BGE_RXLSMODE_ENABLE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_CLRBIT(sc, BGE_RDBDI_MODE, BGE_RBDIMODE_ENABLE);
BGE_CLRBIT(sc, BGE_RDC_MODE, BGE_RDCMODE_ENABLE);
BGE_CLRBIT(sc, BGE_RBDC_MODE, BGE_RBDCMODE_ENABLE);
/*
* Disable all of the transmit blocks
*/
BGE_CLRBIT(sc, BGE_SRS_MODE, BGE_SRSMODE_ENABLE);
BGE_CLRBIT(sc, BGE_SBDI_MODE, BGE_SBDIMODE_ENABLE);
BGE_CLRBIT(sc, BGE_SDI_MODE, BGE_SDIMODE_ENABLE);
BGE_CLRBIT(sc, BGE_RDMA_MODE, BGE_RDMAMODE_ENABLE);
BGE_CLRBIT(sc, BGE_SDC_MODE, BGE_SDCMODE_ENABLE);
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_CLRBIT(sc, BGE_DMAC_MODE, BGE_DMACMODE_ENABLE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_CLRBIT(sc, BGE_SBDC_MODE, BGE_SBDCMODE_ENABLE);
/*
* Shut down all of the memory managers and related
* state machines.
*/
BGE_CLRBIT(sc, BGE_HCC_MODE, BGE_HCCMODE_ENABLE);
BGE_CLRBIT(sc, BGE_WDMA_MODE, BGE_WDMAMODE_ENABLE);
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_CLRBIT(sc, BGE_MBCF_MODE, BGE_MBCFMODE_ENABLE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_FTQ_RESET, 0xFFFFFFFF);
CSR_WRITE_4(sc, BGE_FTQ_RESET, 0);
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750) {
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_CLRBIT(sc, BGE_BMAN_MODE, BGE_BMANMODE_ENABLE);
BGE_CLRBIT(sc, BGE_MARB_MODE, BGE_MARBMODE_ENABLE);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Disable host interrupts. */
BGE_SETBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 1);
/*
* Tell firmware we're shutting down.
*/
BGE_CLRBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
/* Free the RX lists. */
bge_free_rx_ring_std(sc);
/* Free jumbo RX list. */
Fix the statements for checking if we're dealing with a 5705/5750. Pointhat to: me
2004-09-25 05:07:20 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5750)
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
bge_free_rx_ring_jumbo(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Free TX buffers. */
bge_free_tx_ring(sc);
/*
* Isolate/power down the PHY, but leave the media selection
* unchanged so that things will be put back to normal when
* we bring the interface back up.
*/
if (!sc->bge_tbi) {
itmp = ifp->if_flags;
ifp->if_flags |= IFF_UP;
When bge_stop() is called from bge_detach(), mii is already NULL, so check before trying to use it. This fix allows to 'kldunload if_bge' without panicing. Reviewed by: bz MFC after: 3 days
2005-09-04 06:35:59 +00:00
/*
* If we are called from bge_detach(), mii is already NULL.
*/
if (mii != NULL) {
ifm = mii->mii_media.ifm_cur;
mtmp = ifm->ifm_media;
ifm->ifm_media = IFM_ETHER|IFM_NONE;
mii_mediachg(mii);
ifm->ifm_media = mtmp;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_flags = itmp;
}
sc->bge_tx_saved_considx = BGE_TXCONS_UNSET;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
/*
* Stop all chip I/O so that the kernel's probe routines don't
* get confused by errant DMAs when rebooting.
*/
static void
bge_shutdown(dev)
device_t dev;
{
struct bge_softc *sc;
sc = device_get_softc(dev);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bge_stop(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_reset(sc);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
Implement suspend/resume methods to be more ACPI friendly. I'm able to suspend/resume my laptop without this change, but then I need to wait for the watchdog to reset the card. With this change, it is ready immediately. Glanced at by: glebius
2005-09-28 19:20:49 +00:00
static int
bge_suspend(device_t dev)
{
struct bge_softc *sc;
sc = device_get_softc(dev);
BGE_LOCK(sc);
bge_stop(sc);
BGE_UNLOCK(sc);
return (0);
}
static int
bge_resume(device_t dev)
{
struct bge_softc *sc;
struct ifnet *ifp;
sc = device_get_softc(dev);
BGE_LOCK(sc);
ifp = sc->bge_ifp;
if (ifp->if_flags & IFF_UP) {
bge_init_locked(sc);
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
bge_start_locked(ifp);
}
BGE_UNLOCK(sc);
return (0);
}
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
static void
bge_link_upd(sc)
struct bge_softc *sc;
{
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
struct mii_data *mii;
uint32_t link, status;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
BGE_LOCK_ASSERT(sc);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
/* Clear 'pending link event' flag */
sc->bge_link_evt = 0;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
/*
* Process link state changes.
* Grrr. The link status word in the status block does
* not work correctly on the BCM5700 rev AX and BX chips,
* according to all available information. Hence, we have
* to enable MII interrupts in order to properly obtain
* async link changes. Unfortunately, this also means that
* we have to read the MAC status register to detect link
* changes, thereby adding an additional register access to
* the interrupt handler.
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
*
* XXX: perhaps link state detection procedure used for
* BGE_CHIPID_BCM5700_B1 can be used for others BCM5700 revisions.
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
*/
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
sc->bge_chipid != BGE_CHIPID_BCM5700_B1) {
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
status = CSR_READ_4(sc, BGE_MAC_STS);
if (status & BGE_MACSTAT_MI_INTERRUPT) {
callout_stop(&sc->bge_stat_ch);
bge_tick_locked(sc);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
mii = device_get_softc(sc->bge_miibus);
if (!sc->bge_link &&
mii->mii_media_status & IFM_ACTIVE &&
IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
sc->bge_link++;
if (bootverbose)
if_printf(sc->bge_ifp, "link UP\n");
} else if (sc->bge_link &&
(!(mii->mii_media_status & IFM_ACTIVE) ||
IFM_SUBTYPE(mii->mii_media_active) == IFM_NONE)) {
sc->bge_link = 0;
if (bootverbose)
if_printf(sc->bge_ifp, "link DOWN\n");
}
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
/* Clear the interrupt */
CSR_WRITE_4(sc, BGE_MAC_EVT_ENB,
BGE_EVTENB_MI_INTERRUPT);
bge_miibus_readreg(sc->bge_dev, 1, BRGPHY_MII_ISR);
bge_miibus_writereg(sc->bge_dev, 1, BRGPHY_MII_IMR,
BRGPHY_INTRS);
}
return;
}
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if (sc->bge_tbi) {
status = CSR_READ_4(sc, BGE_MAC_STS);
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
if (status & BGE_MACSTAT_TBI_PCS_SYNCHED) {
if (!sc->bge_link) {
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
sc->bge_link++;
if (sc->bge_asicrev == BGE_ASICREV_BCM5704)
BGE_CLRBIT(sc, BGE_MAC_MODE,
BGE_MACMODE_TBI_SEND_CFGS);
CSR_WRITE_4(sc, BGE_MAC_STS, 0xFFFFFFFF);
if (bootverbose)
if_printf(sc->bge_ifp, "link UP\n");
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
if_link_state_change(sc->bge_ifp, LINK_STATE_UP);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
}
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
} else if (sc->bge_link) {
sc->bge_link = 0;
if (bootverbose)
if_printf(sc->bge_ifp, "link DOWN\n");
if_link_state_change(sc->bge_ifp, LINK_STATE_DOWN);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
}
} else {
/*
* Some broken BCM chips have BGE_STATFLAG_LINKSTATE_CHANGED bit
* in status word always set. Workaround this bug by reading
* PHY link status directly.
*/
link = (CSR_READ_4(sc, BGE_MI_STS) & BGE_MISTS_LINK) ? 1 : 0;
if (link != sc->bge_link ||
sc->bge_asicrev == BGE_ASICREV_BCM5700) {
callout_stop(&sc->bge_stat_ch);
bge_tick_locked(sc);
mii = device_get_softc(sc->bge_miibus);
if (!sc->bge_link &&
mii->mii_media_status & IFM_ACTIVE &&
IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
sc->bge_link++;
if (bootverbose)
if_printf(sc->bge_ifp, "link UP\n");
} else if (sc->bge_link &&
(!(mii->mii_media_status & IFM_ACTIVE) ||
IFM_SUBTYPE(mii->mii_media_active) == IFM_NONE)) {
sc->bge_link = 0;
if (bootverbose)
if_printf(sc->bge_ifp, "link DOWN\n");
}
}
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
}
/* Clear the interrupt */
CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED|
BGE_MACSTAT_CFG_CHANGED|BGE_MACSTAT_MI_COMPLETE|
BGE_MACSTAT_LINK_CHANGED);
}
Reference in New Issue Copy Permalink