d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
e33662c6c4
freebsd-dev/sys/dev/bge/if_bge.c

3747 lines
99 KiB
C
Raw Normal View History

Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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.
*
* 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.
*/
#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 <sys/queue.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_memio.h>
#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>
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)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
/* "controller miibus0" required. See GENERIC if you get errors here. */
#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" },
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" },
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 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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 void bge_dma_map_tx_desc (void *, bus_dma_segment_t *, int,
bus_size_t, 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 void bge_handle_events (struct bge_softc *);
static int bge_alloc_jumbo_mem (struct bge_softc *);
static void bge_free_jumbo_mem (struct bge_softc *);
static void *bge_jalloc (struct bge_softc *);
catch up with ext_free prototype change.
2002-06-29 01:36:59 +00:00
static void bge_jfree (void *, void *);
Remove __P.
2002-03-20 02:08:01 +00:00
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 *);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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),
/* 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;
}
/*
* Map an mbuf chain into an TX ring.
*/
static void
bge_dma_map_tx_desc(arg, segs, nseg, mapsize, error)
void *arg;
bus_dma_segment_t *segs;
int nseg;
bus_size_t mapsize;
int error;
{
struct bge_dmamap_arg *ctx;
struct bge_tx_bd *d = NULL;
int i = 0, idx;
if (error)
return;
ctx = arg;
/* Signal error to caller if there's too many segments */
if (nseg > ctx->bge_maxsegs) {
ctx->bge_maxsegs = 0;
return;
}
idx = ctx->bge_idx;
while(1) {
d = &ctx->bge_ring[idx];
d->bge_addr.bge_addr_lo =
htole32(BGE_ADDR_LO(segs[i].ds_addr));
d->bge_addr.bge_addr_hi =
htole32(BGE_ADDR_HI(segs[i].ds_addr));
d->bge_len = htole16(segs[i].ds_len);
d->bge_flags = htole16(ctx->bge_flags);
i++;
if (i == nseg)
break;
BGE_INC(idx, BGE_TX_RING_CNT);
}
d->bge_flags |= htole16(BGE_TXBDFLAG_END);
ctx->bge_maxsegs = nseg;
ctx->bge_idx = idx;
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) {
printf("bge%d: VPD read timed out\n", sc->bge_unit);
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) {
printf("bge%d: bad VPD resource id: expected %x got %x\n",
sc->bge_unit, VPD_RES_ID, res.vr_id);
return;
}
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) {
printf("bge%d: bad VPD resource id: expected %x got %x\n",
sc->bge_unit, VPD_RES_READ, res.vr_id);
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) {
printf("bge%d: eeprom read timed out\n", sc->bge_unit);
return(0);
}
/* Get result. */
byte = CSR_READ_4(sc, BGE_EE_DATA);
*dest = (byte >> ((addr % 4) * 8)) & 0xFF;
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) {
printf("bge%d: PHY read timed out\n", sc->bge_unit);
- 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) {
printf("bge%d: PHY read timed out\n", sc->bge_unit);
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;
}
/*
* Handle events that have triggered interrupts.
*/
static void
bge_handle_events(sc)
struct bge_softc *sc;
{
return;
}
/*
* Memory management for jumbo frames.
*/
static int
bge_alloc_jumbo_mem(sc)
struct bge_softc *sc;
{
caddr_t ptr;
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
register int 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
struct bge_jpool_entry *entry;
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 buffer block */
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_JMEM, 1, BGE_JMEM, 0, NULL, NULL,
&sc->bge_cdata.bge_jumbo_tag);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
if (error) {
printf("bge%d: could not allocate jumbo dma tag\n",
sc->bge_unit);
return (ENOMEM);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
/* Allocate DMA'able memory for jumbo buffer block */
error = bus_dmamem_alloc(sc->bge_cdata.bge_jumbo_tag,
(void **)&sc->bge_ldata.bge_jumbo_buf, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_jumbo_map);
if (error)
return (ENOMEM);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
SLIST_INIT(&sc->bge_jfree_listhead);
SLIST_INIT(&sc->bge_jinuse_listhead);
/*
* Now divide it up into 9K pieces and save the addresses
* in an array.
*/
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
ptr = sc->bge_ldata.bge_jumbo_buf;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_JSLOTS; i++) {
sc->bge_cdata.bge_jslots[i] = ptr;
ptr += BGE_JLEN;
entry = malloc(sizeof(struct bge_jpool_entry),
M_DEVBUF, M_NOWAIT);
if (entry == 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
bge_free_jumbo_mem(sc);
sc->bge_ldata.bge_jumbo_buf = 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
printf("bge%d: no memory for jumbo "
"buffer queue!\n", sc->bge_unit);
return(ENOBUFS);
}
entry->slot = i;
SLIST_INSERT_HEAD(&sc->bge_jfree_listhead,
entry, jpool_entries);
}
return(0);
}
static void
bge_free_jumbo_mem(sc)
struct bge_softc *sc;
{
int i;
struct bge_jpool_entry *entry;
for (i = 0; i < BGE_JSLOTS; i++) {
entry = SLIST_FIRST(&sc->bge_jfree_listhead);
SLIST_REMOVE_HEAD(&sc->bge_jfree_listhead, jpool_entries);
free(entry, M_DEVBUF);
}
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
/* Destroy jumbo buffer block */
if (sc->bge_ldata.bge_rx_jumbo_ring)
bus_dmamem_free(sc->bge_cdata.bge_jumbo_tag,
sc->bge_ldata.bge_jumbo_buf,
sc->bge_cdata.bge_jumbo_map);
if (sc->bge_cdata.bge_rx_jumbo_ring_map)
bus_dmamap_destroy(sc->bge_cdata.bge_jumbo_tag,
sc->bge_cdata.bge_jumbo_map);
if (sc->bge_cdata.bge_jumbo_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_jumbo_tag);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
}
/*
* Allocate a jumbo buffer.
*/
static void *
bge_jalloc(sc)
struct bge_softc *sc;
{
struct bge_jpool_entry *entry;
entry = SLIST_FIRST(&sc->bge_jfree_listhead);
if (entry == NULL) {
printf("bge%d: no free jumbo buffers\n", sc->bge_unit);
return(NULL);
}
SLIST_REMOVE_HEAD(&sc->bge_jfree_listhead, jpool_entries);
SLIST_INSERT_HEAD(&sc->bge_jinuse_listhead, entry, jpool_entries);
return(sc->bge_cdata.bge_jslots[entry->slot]);
}
/*
* Release a jumbo buffer.
*/
static void
bge_jfree(buf, args)
catch up with ext_free prototype change.
2002-06-29 01:36:59 +00:00
void *buf;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
void *args;
{
struct bge_jpool_entry *entry;
struct bge_softc *sc;
int i;
/* Extract the softc struct pointer. */
sc = (struct bge_softc *)args;
if (sc == NULL)
panic("bge_jfree: can't find softc pointer!");
/* calculate the slot this buffer belongs to */
i = ((vm_offset_t)buf
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
- (vm_offset_t)sc->bge_ldata.bge_jumbo_buf) / BGE_JLEN;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 < 0) || (i >= BGE_JSLOTS))
panic("bge_jfree: asked to free buffer that we don't manage!");
entry = SLIST_FIRST(&sc->bge_jinuse_listhead);
if (entry == NULL)
panic("bge_jfree: buffer not in use!");
entry->slot = i;
SLIST_REMOVE_HEAD(&sc->bge_jinuse_listhead, jpool_entries);
SLIST_INSERT_HEAD(&sc->bge_jfree_listhead, entry, jpool_entries);
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) {
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_new == NULL) {
return(ENOBUFS);
}
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
MCLGET(m_new, M_DONTWAIT);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_new->m_flags & M_EXT)) {
m_freem(m_new);
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 (m == NULL)
m_freem(m_new);
return(ENOMEM);
}
r->bge_addr.bge_addr_lo = htole32(BGE_ADDR_LO(ctx.bge_busaddr));
r->bge_addr.bge_addr_hi = htole32(BGE_ADDR_HI(ctx.bge_busaddr));
r->bge_flags = htole16(BGE_RXBDFLAG_END);
r->bge_len = htole16(m_new->m_len);
r->bge_idx = htole16(i);
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;
{
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) {
caddr_t *buf = NULL;
/* Allocate the mbuf. */
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_new == NULL) {
return(ENOBUFS);
}
/* Allocate the jumbo buffer */
buf = bge_jalloc(sc);
if (buf == NULL) {
m_freem(m_new);
printf("bge%d: jumbo allocation failed "
"-- packet dropped!\n", sc->bge_unit);
return(ENOBUFS);
}
/* Attach the buffer to the mbuf. */
m_new->m_data = (void *) buf;
m_new->m_len = m_new->m_pkthdr.len = BGE_JUMBO_FRAMELEN;
MEXTADD(m_new, buf, BGE_JUMBO_FRAMELEN, bge_jfree,
(struct bge_softc *)sc, 0, EXT_NET_DRV);
} else {
m_new = m;
m_new->m_data = m_new->m_ext.ext_buf;
m_new->m_ext.ext_size = BGE_JUMBO_FRAMELEN;
}
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
/* Set up the descriptor. */
sc->bge_cdata.bge_rx_jumbo_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_jumbo_ring[i];
ctx.bge_maxsegs = 1;
ctx.sc = sc;
error = bus_dmamap_load(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_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 (m == NULL)
m_freem(m_new);
return(ENOMEM);
}
r->bge_addr.bge_addr_lo = htole32(BGE_ADDR_LO(ctx.bge_busaddr));
r->bge_addr.bge_addr_hi = htole32(BGE_ADDR_HI(ctx.bge_busaddr));
r->bge_flags = htole16(BGE_RXBDFLAG_END|BGE_RXBDFLAG_JUMBO_RING);
r->bge_len = htole16(m_new->m_len);
r->bge_idx = htole16(i);
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
return(0);
}
/*
* 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) {
m_freem(sc->bge_cdata.bge_rx_std_chain[i]);
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
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[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
}
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;
{
int i;
struct bge_rcb *rcb;
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;
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
rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(0, 0);
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) {
m_freem(sc->bge_cdata.bge_rx_jumbo_chain[i]);
sc->bge_cdata.bge_rx_jumbo_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
bus_dmamap_unload(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[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
}
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],
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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) {
m_freem(sc->bge_cdata.bge_tx_chain[i]);
sc->bge_cdata.bge_tx_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
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[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
}
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
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_HOST_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_HOST_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
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 = &sc->arpcom.ac_if;
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. */
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);
}
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
/* Set endianness before we access any non-PCI registers. */
#if BYTE_ORDER == BIG_ENDIAN
pci_write_config(sc->bge_dev, BGE_PCI_MISC_CTL,
BGE_BIGENDIAN_INIT, 4);
#else
pci_write_config(sc->bge_dev, BGE_PCI_MISC_CTL,
BGE_LITTLEENDIAN_INIT, 4);
#endif
/*
* 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) {
printf("bge%d: RX CPU self-diagnostics failed!\n",
sc->bge_unit);
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.
*/
CSR_WRITE_4(sc, BGE_MODE_CTL, BGE_MODECTL_WORDSWAP_NONFRAME|
BGE_MODECTL_BYTESWAP_DATA|BGE_MODECTL_WORDSWAP_DATA|
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
* 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
*/
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;
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
volatile struct bge_rcb *vrcb;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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) {
printf("bge%d: buffer manager failed to start\n",
sc->bge_unit);
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) {
printf("bge%d: flow-through queue init failed\n",
sc->bge_unit);
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);
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(BGE_MAX_FRAMELEN,
BGE_RCB_FLAG_RING_DISABLED);
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.
*/
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
vrcb = (volatile struct bge_rcb *)(sc->bge_vhandle + BGE_MEMWIN_START +
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_SEND_RING_RCB);
for (i = 0; i < BGE_TX_RINGS_EXTSSRAM_MAX; i++) {
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
vrcb->bge_maxlen_flags =
BGE_RCB_MAXLEN_FLAGS(0, BGE_RCB_FLAG_RING_DISABLED);
vrcb->bge_nicaddr = 0;
vrcb++;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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) */
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
vrcb = (volatile struct bge_rcb *)(sc->bge_vhandle + BGE_MEMWIN_START +
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_SEND_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
vrcb->bge_hostaddr.bge_addr_lo =
htole32(BGE_ADDR_LO(sc->bge_ldata.bge_tx_ring_paddr));
vrcb->bge_hostaddr.bge_addr_hi =
htole32(BGE_ADDR_HI(sc->bge_ldata.bge_tx_ring_paddr));
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
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 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
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 */
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
vrcb = (volatile struct bge_rcb *)(sc->bge_vhandle + BGE_MEMWIN_START +
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_RX_RETURN_RING_RCB);
for (i = 0; i < BGE_RX_RINGS_MAX; i++) {
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
vrcb->bge_hostaddr.bge_addr_hi = 0;
vrcb->bge_hostaddr.bge_addr_lo = 0;
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,
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
BGE_RCB_FLAG_RING_DISABLED);
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);
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
vrcb++;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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.
*/
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
vrcb = (volatile struct bge_rcb *)(sc->bge_vhandle + BGE_MEMWIN_START +
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_RX_RETURN_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
vrcb->bge_hostaddr.bge_addr_lo =
BGE_ADDR_LO(sc->bge_ldata.bge_rx_return_ring_paddr);
vrcb->bge_hostaddr.bge_addr_hi =
BGE_ADDR_HI(sc->bge_ldata.bge_rx_return_ring_paddr);
bus_dmamap_sync(sc->bge_cdata.bge_rx_return_ring_tag,
sc->bge_cdata.bge_rx_return_ring_map, BUS_DMASYNC_PREWRITE);
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
vrcb->bge_nicaddr = 0x00000000;
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
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,
sc->arpcom.ac_enaddr[0] + sc->arpcom.ac_enaddr[1] +
sc->arpcom.ac_enaddr[2] + sc->arpcom.ac_enaddr[3] +
sc->arpcom.ac_enaddr[4] + sc->arpcom.ac_enaddr[5] +
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) {
printf("bge%d: host coalescing engine failed to idle\n",
sc->bge_unit);
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));
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_PREWRITE);
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);
/* 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);
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
} 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);
- 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_BCM5700)
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
/* 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_unit = device_get_unit(dev);
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 */
if (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,
sc->bge_cdata.bge_rx_std_ring_map);
if (sc->bge_cdata.bge_rx_std_ring_map) {
bus_dmamap_unload(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map);
bus_dmamap_destroy(sc->bge_cdata.bge_rx_std_ring_tag,
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 */
if (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,
sc->bge_cdata.bge_rx_jumbo_ring_map);
if (sc->bge_cdata.bge_rx_jumbo_ring_map) {
bus_dmamap_unload(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map);
bus_dmamap_destroy(sc->bge_cdata.bge_rx_jumbo_ring_tag,
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 */
if (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,
sc->bge_cdata.bge_rx_return_ring_map);
if (sc->bge_cdata.bge_rx_return_ring_map) {
bus_dmamap_unload(sc->bge_cdata.bge_rx_return_ring_tag,
sc->bge_cdata.bge_rx_return_ring_map);
bus_dmamap_destroy(sc->bge_cdata.bge_rx_return_ring_tag,
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 */
if (sc->bge_ldata.bge_tx_ring)
bus_dmamem_free(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_ldata.bge_tx_ring,
sc->bge_cdata.bge_tx_ring_map);
if (sc->bge_cdata.bge_tx_ring_map) {
bus_dmamap_unload(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_cdata.bge_tx_ring_map);
bus_dmamap_destroy(sc->bge_cdata.bge_tx_ring_tag,
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 */
if (sc->bge_ldata.bge_status_block)
bus_dmamem_free(sc->bge_cdata.bge_status_tag,
sc->bge_ldata.bge_status_block,
sc->bge_cdata.bge_status_map);
if (sc->bge_cdata.bge_status_map) {
bus_dmamap_unload(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map);
bus_dmamap_destroy(sc->bge_cdata.bge_status_tag,
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 */
if (sc->bge_ldata.bge_stats)
bus_dmamem_free(sc->bge_cdata.bge_stats_tag,
sc->bge_ldata.bge_stats,
sc->bge_cdata.bge_stats_map);
if (sc->bge_cdata.bge_stats_map) {
bus_dmamap_unload(sc->bge_cdata.bge_stats_tag,
sc->bge_cdata.bge_stats_map);
bus_dmamap_destroy(sc->bge_cdata.bge_stats_tag,
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;
int nseg, i, error;
struct bge_dmamap_arg ctx;
sc = device_get_softc(dev);
/*
* Allocate the parent bus DMA tag appropriate for PCI.
*/
#define BGE_NSEG_NEW 32
error = bus_dma_tag_create(NULL, /* parent */
PAGE_SIZE, 0, /* alignment, boundary */
BUS_SPACE_MAXADDR, /* lowaddr */
BUS_SPACE_MAXADDR_32BIT,/* highaddr */
NULL, NULL, /* filter, filterarg */
MAXBSIZE, BGE_NSEG_NEW, /* maxsize, nsegments */
BUS_SPACE_MAXSIZE_32BIT,/* maxsegsize */
BUS_DMA_ALLOCNOW, /* flags */
NULL, NULL, /* lockfunc, lockarg */
&sc->bge_cdata.bge_parent_tag);
/*
* Create tag for RX mbufs.
*/
nseg = 32;
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag, ETHER_ALIGN,
0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, MCLBYTES * nseg, nseg, MCLBYTES, 0, NULL, NULL,
&sc->bge_cdata.bge_mtag);
if (error) {
device_printf(dev, "could not allocate dma tag\n");
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) {
device_printf(dev, "can't create DMA map for RX\n");
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) {
device_printf(dev, "can't create DMA map for RX\n");
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) {
device_printf(dev, "could not allocate dma tag\n");
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);
if (error)
return (ENOMEM);
bzero((char *)sc->bge_ldata.bge_rx_std_ring, BGE_STD_RX_RING_SZ);
/* 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,
ETHER_ALIGN, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, MCLBYTES * nseg, nseg, BGE_JLEN, 0, NULL, NULL,
&sc->bge_cdata.bge_mtag_jumbo);
if (error) {
device_printf(dev, "could not allocate dma tag\n");
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) {
device_printf(dev, "could not allocate dma tag\n");
return (ENOMEM);
}
/* Allocate DMA'able memory for jumbo RX ring */
error = bus_dmamem_alloc(sc->bge_cdata.bge_rx_jumbo_ring_tag,
(void **)&sc->bge_ldata.bge_rx_jumbo_ring, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_rx_jumbo_ring_map);
if (error)
return (ENOMEM);
bzero((char *)sc->bge_ldata.bge_rx_jumbo_ring,
BGE_JUMBO_RX_RING_SZ);
/* 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) {
device_printf(dev,
"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) {
device_printf(dev, "could not allocate dma tag\n");
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);
if (error)
return (ENOMEM);
bzero((char *)sc->bge_ldata.bge_rx_return_ring,
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) {
device_printf(dev, "could not allocate dma tag\n");
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);
if (error)
return (ENOMEM);
bzero((char *)sc->bge_ldata.bge_tx_ring, BGE_TX_RING_SZ);
/* 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) {
device_printf(dev, "could not allocate dma tag\n");
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);
if (error)
return (ENOMEM);
bzero((char *)sc->bge_ldata.bge_status_block, BGE_STATUS_BLK_SZ);
/* 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,
NULL, BGE_STATS_SZ, 1, BGE_STATS_SZ, 0, NULL, NULL,
&sc->bge_cdata.bge_stats_tag);
if (error) {
device_printf(dev, "could not allocate dma tag\n");
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);
if (error)
return (ENOMEM);
bzero((char *)sc->bge_ldata.bge_stats, BGE_STATS_SZ);
/* 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;
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
u_int32_t mac_addr = 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
int unit, error = 0, rid;
sc = device_get_softc(dev);
unit = device_get_unit(dev);
sc->bge_dev = dev;
sc->bge_unit = unit;
/*
* 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) {
printf ("bge%d: couldn't map memory\n", unit);
error = ENXIO;
goto fail;
}
sc->bge_btag = rman_get_bustag(sc->bge_res);
sc->bge_bhandle = rman_get_bushandle(sc->bge_res);
sc->bge_vhandle = (vm_offset_t)rman_get_virtual(sc->bge_res);
/* Allocate interrupt */
rid = 0;
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->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) {
printf("bge%d: couldn't map interrupt\n", unit);
error = ENXIO;
goto fail;
}
sc->bge_unit = unit;
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);
/*
* 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)) {
printf("bge%d: chip initialization failed\n", sc->bge_unit);
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
/*
* Get station address from the EEPROM.
*/
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
mac_addr = bge_readmem_ind(sc, 0x0c14);
if ((mac_addr >> 16) == 0x484b) {
sc->arpcom.ac_enaddr[0] = (u_char)(mac_addr >> 8);
sc->arpcom.ac_enaddr[1] = (u_char)mac_addr;
mac_addr = bge_readmem_ind(sc, 0x0c18);
sc->arpcom.ac_enaddr[2] = (u_char)(mac_addr >> 24);
sc->arpcom.ac_enaddr[3] = (u_char)(mac_addr >> 16);
sc->arpcom.ac_enaddr[4] = (u_char)(mac_addr >> 8);
sc->arpcom.ac_enaddr[5] = (u_char)mac_addr;
} else if (bge_read_eeprom(sc, (caddr_t)&sc->arpcom.ac_enaddr,
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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)) {
printf("bge%d: failed to read station address\n", unit);
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)) {
printf ("bge%d: failed to allocate DMA resources\n",
sc->bge_unit);
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
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
/*
* Try to allocate memory for jumbo buffers.
* The 5705 does not appear to support jumbo frames.
*/
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
if (bge_alloc_jumbo_mem(sc)) {
printf("bge%d: jumbo buffer allocation "
"failed\n", sc->bge_unit);
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 */
ifp = &sc->arpcom.ac_if;
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;
ifp->if_snd.ifq_maxlen = BGE_TX_RING_CNT - 1;
ifp->if_hwassist = BGE_CSUM_FEATURES;
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
/* NB: the code for RX csum offload is disabled for now */
ifp->if_capabilities = IFCAP_TXCSUM | IFCAP_VLAN_HWTAGGING |
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
IFCAP_VLAN_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
ifp->if_capenable = ifp->if_capabilities;
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 {
bge_read_eeprom(sc, (caddr_t)&hwcfg,
BGE_EE_HWCFG_OFFSET, sizeof(hwcfg));
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)) {
printf("bge%d: MII without any PHY!\n", sc->bge_unit);
bge_release_resources(sc);
bge_free_jumbo_mem(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.
*/
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
ether_ifattach(ifp, sc->arpcom.ac_enaddr);
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) {
bge_release_resources(sc);
printf("bge%d: couldn't set up irq\n", unit);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
ifp = &sc->arpcom.ac_if;
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);
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_jumbo_mem(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(0);
}
static void
bge_release_resources(sc)
struct bge_softc *sc;
{
device_t dev;
dev = sc->bge_dev;
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);
if (sc->bge_intrhand != NULL)
bus_teardown_intr(dev, sc->bge_irq, sc->bge_intrhand);
if (sc->bge_irq != NULL)
bus_release_resource(dev, SYS_RES_IRQ, 0, sc->bge_irq);
if (sc->bge_res != NULL)
bus_release_resource(dev, SYS_RES_MEMORY,
BGE_PCI_BAR0, sc->bge_res);
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_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|
BGE_PCIMISCCTL_ENDIAN_WORDSWAP|BGE_PCIMISCCTL_PCISTATE_RW, 4);
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);
}
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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|
BGE_PCIMISCCTL_ENDIAN_WORDSWAP|BGE_PCIMISCCTL_PCISTATE_RW, 4);
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);
}
if (i == BGE_TIMEOUT) {
printf("bge%d: firmware handshake timed out\n", sc->bge_unit);
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 */
CSR_WRITE_4(sc, BGE_MODE_CTL, BGE_MODECTL_BYTESWAP_NONFRAME|
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:
* 1) the frame is from the jumbo recieve ring
* 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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 = &sc->arpcom.ac_if;
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,
sc->bge_cdata.bge_rx_return_ring_map, BUS_DMASYNC_POSTWRITE);
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;
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++;
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
#ifndef __i386__
/*
* The i386 allows unaligned accesses, but for other
* platforms we must make sure the payload is aligned.
*/
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;
Disabled input hardware checksum due to it being calculated incorrected for some packets, in particular small (0 byte payload) packets. May also be related to TCP options. Approved by: wpaul MFC after: 3 days
2001-12-14 22:04:58 +00:00
#if 0 /* currently broken for some packets, possibly related to TCP options */
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) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_flags |= CSUM_IP_CHECKED;
if ((cur_rx->bge_ip_csum ^ 0xffff) == 0)
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
if (cur_rx->bge_flags & BGE_RXBDFLAG_TCP_UDP_CSUM) {
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
}
}
Disabled input hardware checksum due to it being calculated incorrected for some packets, in particular small (0 byte payload) packets. May also be related to TCP options. Approved by: wpaul MFC after: 3 days
2001-12-14 22:04:58 +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
/*
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
*/
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 (have_tag)
VLAN_INPUT_TAG(ifp, m, vlan_tag, 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
}
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,
sc->bge_cdata.bge_rx_return_ring_map, BUS_DMASYNC_PREWRITE);
bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map,
BUS_DMASYNC_POSTREAD|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) {
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
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 = &sc->arpcom.ac_if;
/*
* 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) {
m_freem(sc->bge_cdata.bge_tx_chain[idx]);
sc->bge_cdata.bge_tx_chain[idx] = 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
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[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
}
sc->bge_txcnt--;
BGE_INC(sc->bge_tx_saved_considx, BGE_TX_RING_CNT);
ifp->if_timer = 0;
}
if (cur_tx != NULL)
ifp->if_flags &= ~IFF_OACTIVE;
return;
}
static void
bge_intr(xsc)
void *xsc;
{
struct bge_softc *sc;
struct ifnet *ifp;
- 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
u_int32_t statusword;
Commit patch to supress spurious link change events. Apparently, with copper NICs, a link change event is posted whenever MII autopolling is toggled off and on, which happens whenever someone calls bge_miibus_readreg() or bge_miibus_writereg() to access the PHY registers. This means anytime someone called the SIOCGIFMEDIA ioctl on a bge interface, the link would reset. Even a simple "ifconfig bge0" would do it, though other apps like dhclient or the PPPoE daemon could trigger it as well. An obvious symptom of this problem is lots of "bgeX: gigabit link up" messages appearing on the console for no apparent reason. Through experimentation, I determined that when a real link change event occurs, the BGE_MIMODE_AUTOPOLL in the BGE_MI_MODE register is always set, so now if we have a copper NIC and an link change event occurs and the BGE_MIMODE_AUTOPOLL bit is clear, we ignore the event. Note that this does not apply to the original BCM5700 chip since we use a different method for sensing link changes with that chip (the status block method was broken), nor to fiber optic NICs since they don't use the GMII PHY access registers.
2004-09-17 04:58:17 +00:00
u_int32_t status, mimode;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
ifp = &sc->arpcom.ac_if;
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);
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,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_POSTWRITE);
- 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
#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);
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
/*
* 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,
Fix typo in a comment
2004-01-13 11:31:09 +00:00
* according to all available information. Hence, we have
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
* 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.
*/
- 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_BCM5700) {
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
status = CSR_READ_4(sc, BGE_MAC_STS);
if (status & BGE_MACSTAT_MI_INTERRUPT) {
sc->bge_link = 0;
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);
bge_tick_locked(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
/* 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);
}
} else {
- 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
if (statusword & BGE_STATFLAG_LINKSTATE_CHANGED) {
Squelch spurious "gigabit link up" messages generated on some fiber NICs (mainly the 3Com 3c996B/BCM5701). For some reason that I don't fully understand, the 5701 signals PCS encoding errors as though they were link change events, i.e. the 'link state changed' bit in the status word of the status block is updated and an interrupt is generated. This would cause the bge_tick() function to be invoked and a "gigabit link up" message to be printed on the console. To avoid this, the interrupt handler now checks the MAC status register when a link change interrupt is triggered, and it will only call the bge_tick() function if the 'PCS encoding error detected' bit is clear. (This change should have no effect on copper NICs since this bit can only ever be set in TBI mode. I do not know how it affects 5704 NICs with a BCM8002 SERDES PHY.) Special thanks to: Sherry Rogers at UCB for allowing me access to one of their traffic monitor boxes so I could diagnose this problem.
2003-07-11 08:19:52 +00:00
/*
* Sometimes PCS encoding errors are detected in
* TBI mode (on fiber NICs), and for some reason
* the chip will signal them as link changes.
* If we get a link change event, but the 'PCS
* encoding error' bit in the MAC status register
* is set, don't bother doing a link check.
* This avoids spurious "gigabit link up" messages
* that sometimes appear on fiber NICs during
* periods of heavy traffic. (There should be no
* effect on copper NICs.)
Commit patch to supress spurious link change events. Apparently, with copper NICs, a link change event is posted whenever MII autopolling is toggled off and on, which happens whenever someone calls bge_miibus_readreg() or bge_miibus_writereg() to access the PHY registers. This means anytime someone called the SIOCGIFMEDIA ioctl on a bge interface, the link would reset. Even a simple "ifconfig bge0" would do it, though other apps like dhclient or the PPPoE daemon could trigger it as well. An obvious symptom of this problem is lots of "bgeX: gigabit link up" messages appearing on the console for no apparent reason. Through experimentation, I determined that when a real link change event occurs, the BGE_MIMODE_AUTOPOLL in the BGE_MI_MODE register is always set, so now if we have a copper NIC and an link change event occurs and the BGE_MIMODE_AUTOPOLL bit is clear, we ignore the event. Note that this does not apply to the original BCM5700 chip since we use a different method for sensing link changes with that chip (the status block method was broken), nor to fiber optic NICs since they don't use the GMII PHY access registers.
2004-09-17 04:58:17 +00:00
*
* If we do have a copper NIC (bge_tbi == 0) then
* check that the AUTOPOLL bit is set before
* processing the event as a real link change.
* Turning AUTOPOLL on and off in the MII read/write
* functions will often trigger a link status
* interrupt for no reason.
Squelch spurious "gigabit link up" messages generated on some fiber NICs (mainly the 3Com 3c996B/BCM5701). For some reason that I don't fully understand, the 5701 signals PCS encoding errors as though they were link change events, i.e. the 'link state changed' bit in the status word of the status block is updated and an interrupt is generated. This would cause the bge_tick() function to be invoked and a "gigabit link up" message to be printed on the console. To avoid this, the interrupt handler now checks the MAC status register when a link change interrupt is triggered, and it will only call the bge_tick() function if the 'PCS encoding error detected' bit is clear. (This change should have no effect on copper NICs since this bit can only ever be set in TBI mode. I do not know how it affects 5704 NICs with a BCM8002 SERDES PHY.) Special thanks to: Sherry Rogers at UCB for allowing me access to one of their traffic monitor boxes so I could diagnose this problem.
2003-07-11 08:19:52 +00:00
*/
status = CSR_READ_4(sc, BGE_MAC_STS);
Commit patch to supress spurious link change events. Apparently, with copper NICs, a link change event is posted whenever MII autopolling is toggled off and on, which happens whenever someone calls bge_miibus_readreg() or bge_miibus_writereg() to access the PHY registers. This means anytime someone called the SIOCGIFMEDIA ioctl on a bge interface, the link would reset. Even a simple "ifconfig bge0" would do it, though other apps like dhclient or the PPPoE daemon could trigger it as well. An obvious symptom of this problem is lots of "bgeX: gigabit link up" messages appearing on the console for no apparent reason. Through experimentation, I determined that when a real link change event occurs, the BGE_MIMODE_AUTOPOLL in the BGE_MI_MODE register is always set, so now if we have a copper NIC and an link change event occurs and the BGE_MIMODE_AUTOPOLL bit is clear, we ignore the event. Note that this does not apply to the original BCM5700 chip since we use a different method for sensing link changes with that chip (the status block method was broken), nor to fiber optic NICs since they don't use the GMII PHY access registers.
2004-09-17 04:58:17 +00:00
mimode = CSR_READ_4(sc, BGE_MI_MODE);
Silence unwanted 'gigabit link up' messages: do not treat the BGE_MACSTAT_MI_COMPLETE bit in the MAC status register as a link change indicator. We turn this bit on now because some of the newer chips need it, but it usually just means that reading/writing an MII/GMII register has completed, not that a link change has occured.
2003-08-10 01:32:16 +00:00
if (!(status & (BGE_MACSTAT_PORT_DECODE_ERROR|
Commit patch to supress spurious link change events. Apparently, with copper NICs, a link change event is posted whenever MII autopolling is toggled off and on, which happens whenever someone calls bge_miibus_readreg() or bge_miibus_writereg() to access the PHY registers. This means anytime someone called the SIOCGIFMEDIA ioctl on a bge interface, the link would reset. Even a simple "ifconfig bge0" would do it, though other apps like dhclient or the PPPoE daemon could trigger it as well. An obvious symptom of this problem is lots of "bgeX: gigabit link up" messages appearing on the console for no apparent reason. Through experimentation, I determined that when a real link change event occurs, the BGE_MIMODE_AUTOPOLL in the BGE_MI_MODE register is always set, so now if we have a copper NIC and an link change event occurs and the BGE_MIMODE_AUTOPOLL bit is clear, we ignore the event. Note that this does not apply to the original BCM5700 chip since we use a different method for sensing link changes with that chip (the status block method was broken), nor to fiber optic NICs since they don't use the GMII PHY access registers.
2004-09-17 04:58:17 +00:00
BGE_MACSTAT_MI_COMPLETE)) && (!sc->bge_tbi &&
(mimode & BGE_MIMODE_AUTOPOLL))) {
Squelch spurious "gigabit link up" messages generated on some fiber NICs (mainly the 3Com 3c996B/BCM5701). For some reason that I don't fully understand, the 5701 signals PCS encoding errors as though they were link change events, i.e. the 'link state changed' bit in the status word of the status block is updated and an interrupt is generated. This would cause the bge_tick() function to be invoked and a "gigabit link up" message to be printed on the console. To avoid this, the interrupt handler now checks the MAC status register when a link change interrupt is triggered, and it will only call the bge_tick() function if the 'PCS encoding error detected' bit is clear. (This change should have no effect on copper NICs since this bit can only ever be set in TBI mode. I do not know how it affects 5704 NICs with a BCM8002 SERDES PHY.) Special thanks to: Sherry Rogers at UCB for allowing me access to one of their traffic monitor boxes so I could diagnose this problem.
2003-07-11 08:19:52 +00:00
sc->bge_link = 0;
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);
bge_tick_locked(sc);
Squelch spurious "gigabit link up" messages generated on some fiber NICs (mainly the 3Com 3c996B/BCM5701). For some reason that I don't fully understand, the 5701 signals PCS encoding errors as though they were link change events, i.e. the 'link state changed' bit in the status word of the status block is updated and an interrupt is generated. This would cause the bge_tick() function to be invoked and a "gigabit link up" message to be printed on the console. To avoid this, the interrupt handler now checks the MAC status register when a link change interrupt is triggered, and it will only call the bge_tick() function if the 'PCS encoding error detected' bit is clear. (This change should have no effect on copper NICs since this bit can only ever be set in TBI mode. I do not know how it affects 5704 NICs with a BCM8002 SERDES PHY.) Special thanks to: Sherry Rogers at UCB for allowing me access to one of their traffic monitor boxes so I could diagnose this problem.
2003-07-11 08:19:52 +00:00
}
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
/* Clear the interrupt */
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);
- 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
/* Force flush the status block cached by PCI bridge */
CSR_READ_4(sc, BGE_MBX_IRQ0_LO);
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
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_RUNNING) {
/* Check RX return ring producer/consumer */
bge_rxeof(sc);
/* Check TX ring producer/consumer */
bge_txeof(sc);
}
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,
sc->bge_cdata.bge_status_map, 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
bge_handle_events(sc);
/* Re-enable interrupts. */
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 0);
if (ifp->if_flags & IFF_RUNNING && ifp->if_snd.ifq_head != NULL)
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;
struct ifmedia *ifm = NULL;
struct ifnet *ifp;
ifp = &sc->arpcom.ac_if;
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);
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);
if (sc->bge_link)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
if (sc->bge_tbi) {
ifm = &sc->bge_ifmedia;
if (CSR_READ_4(sc, BGE_MAC_STS) &
BGE_MACSTAT_TBI_PCS_SYNCHED) {
sc->bge_link++;
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
if (sc->bge_asicrev == BGE_ASICREV_BCM5704)
BGE_CLRBIT(sc, BGE_MAC_MODE,
BGE_MACMODE_TBI_SEND_CFGS);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_STS, 0xFFFFFFFF);
printf("bge%d: gigabit link up\n", sc->bge_unit);
if (ifp->if_snd.ifq_head != NULL)
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
}
return;
}
mii = device_get_softc(sc->bge_miibus);
mii_tick(mii);
Do not call mii_polltick() immediately after mii_tick(). Poiinted out by: wpaul
2001-09-29 19:31:29 +00:00
if (!sc->bge_link && mii->mii_media_status & IFM_ACTIVE &&
IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
sc->bge_link++;
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 ||
Do not call mii_polltick() immediately after mii_tick(). Poiinted out by: wpaul
2001-09-29 19:31:29 +00:00
IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_SX)
printf("bge%d: gigabit link up\n",
sc->bge_unit);
if (ifp->if_snd.ifq_head != NULL)
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
}
return;
}
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;
int i;
ifp = &sc->arpcom.ac_if;
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++;
}
ifp->if_collisions +=
(stats.dot3StatsSingleCollisionFrames +
stats.dot3StatsMultipleCollisionFrames +
stats.dot3StatsExcessiveCollisions +
stats.dot3StatsLateCollisions) -
ifp->if_collisions;
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_stats_update(sc)
struct bge_softc *sc;
{
struct ifnet *ifp;
struct bge_stats *stats;
ifp = &sc->arpcom.ac_if;
stats = (struct bge_stats *)(sc->bge_vhandle +
BGE_MEMWIN_START + BGE_STATS_BLOCK);
ifp->if_collisions +=
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
(stats->txstats.dot3StatsSingleCollisionFrames.bge_addr_lo +
stats->txstats.dot3StatsMultipleCollisionFrames.bge_addr_lo +
stats->txstats.dot3StatsExcessiveCollisions.bge_addr_lo +
stats->txstats.dot3StatsLateCollisions.bge_addr_lo) -
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_collisions;
#ifdef notdef
ifp->if_collisions +=
(sc->bge_rdata->bge_info.bge_stats.dot3StatsSingleCollisionFrames +
sc->bge_rdata->bge_info.bge_stats.dot3StatsMultipleCollisionFrames +
sc->bge_rdata->bge_info.bge_stats.dot3StatsExcessiveCollisions +
sc->bge_rdata->bge_info.bge_stats.dot3StatsLateCollisions) -
ifp->if_collisions;
#endif
return;
}
/*
* 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;
u_int32_t *txidx;
{
struct bge_tx_bd *f = NULL;
u_int16_t csum_flags = 0;
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
struct m_tag *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
struct bge_dmamap_arg ctx;
bus_dmamap_t map;
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_head->m_pkthdr.csum_flags) {
if (m_head->m_pkthdr.csum_flags & CSUM_IP)
csum_flags |= BGE_TXBDFLAG_IP_CSUM;
if (m_head->m_pkthdr.csum_flags & (CSUM_TCP | CSUM_UDP))
csum_flags |= BGE_TXBDFLAG_TCP_UDP_CSUM;
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;
}
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
mtag = VLAN_OUTPUT_TAG(&sc->arpcom.ac_if, 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
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
ctx.sc = sc;
ctx.bge_idx = *txidx;
ctx.bge_ring = sc->bge_ldata.bge_tx_ring;
ctx.bge_flags = csum_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
/*
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
*/
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
ctx.bge_maxsegs = (BGE_TX_RING_CNT - sc->bge_txcnt) - 16;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
map = sc->bge_cdata.bge_tx_dmamap[*txidx];
error = bus_dmamap_load_mbuf(sc->bge_cdata.bge_mtag, map,
m_head, bge_dma_map_tx_desc, &ctx, BUS_DMA_NOWAIT);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
if (error || ctx.bge_maxsegs == 0 /*||
ctx.bge_idx == sc->bge_tx_saved_considx*/)
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
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.
*/
sc->bge_cdata.bge_tx_dmamap[*txidx] =
sc->bge_cdata.bge_tx_dmamap[ctx.bge_idx];
sc->bge_cdata.bge_tx_dmamap[ctx.bge_idx] = map;
sc->bge_cdata.bge_tx_chain[ctx.bge_idx] = m_head;
sc->bge_txcnt += ctx.bge_maxsegs;
f = &sc->bge_ldata.bge_tx_ring[*txidx];
if (mtag != NULL) {
f->bge_flags |= htole16(BGE_TXBDFLAG_VLAN_TAG);
f->bge_vlan_tag = htole16(VLAN_TAG_VALUE(mtag));
} else {
f->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
BGE_INC(ctx.bge_idx, BGE_TX_RING_CNT);
*txidx = ctx.bge_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
return(0);
}
/*
* 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;
u_int32_t prodidx = 0;
sc = ifp->if_softc;
if (!sc->bge_link && ifp->if_snd.ifq_len < 10)
return;
prodidx = CSR_READ_4(sc, BGE_MBX_TX_HOST_PROD0_LO);
while(sc->bge_cdata.bge_tx_chain[prodidx] == NULL) {
IF_DEQUEUE(&ifp->if_snd, m_head);
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.
*
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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) {
IF_PREPEND(&ifp->if_snd, m_head);
ifp->if_flags |= IFF_OACTIVE;
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)) {
IF_PREPEND(&ifp->if_snd, m_head);
ifp->if_flags |= IFF_OACTIVE;
break;
}
/*
* 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
}
/* 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
/*
* 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
ifp = &sc->arpcom.ac_if;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
if (ifp->if_flags & IFF_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)) {
printf("bge%d: initialization failure\n", sc->bge_unit);
return;
}
ifp = &sc->arpcom.ac_if;
/* 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. */
m = (u_int16_t *)&sc->arpcom.ac_enaddr[0];
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)
printf ("bge%d: 5705 A0 chip failed to load RX ring\n",
sc->bge_unit);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
/* Enable host interrupts. */
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);
bge_ifmedia_upd(ifp);
ifp->if_flags |= IFF_RUNNING;
ifp->if_flags &= ~IFF_OACTIVE;
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);
return;
}
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:
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 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);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
sc->bge_link = 0;
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)
ifmr->ifm_active |= IFM_HDX;
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;
ifp->if_flags &= ~IFF_RUNNING;
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
* second or two.
*/
if (ifp->if_flags & IFF_RUNNING &&
ifp->if_flags & IFF_PROMISC &&
!(sc->bge_if_flags & IFF_PROMISC)) {
BGE_SETBIT(sc, BGE_RX_MODE,
BGE_RXMODE_RX_PROMISC);
} else if (ifp->if_flags & IFF_RUNNING &&
!(ifp->if_flags & IFF_PROMISC) &&
sc->bge_if_flags & IFF_PROMISC) {
BGE_CLRBIT(sc, BGE_RX_MODE,
BGE_RXMODE_RX_PROMISC);
} 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 {
if (ifp->if_flags & IFF_RUNNING) {
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:
if (ifp->if_flags & IFF_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;
case SIOCSIFCAP:
mask = ifr->ifr_reqcap ^ ifp->if_capenable;
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
/* NB: the code for RX csum offload is disabled for now */
if (mask & IFCAP_TXCSUM) {
ifp->if_capenable ^= IFCAP_TXCSUM;
if (IFCAP_TXCSUM & ifp->if_capenable)
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;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
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;
printf("bge%d: watchdog timeout -- resetting\n", sc->bge_unit);
ifp->if_flags &= ~IFF_RUNNING;
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 = &sc->arpcom.ac_if;
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;
ifm = mii->mii_media.ifm_cur;
mtmp = ifm->ifm_media;
ifm->ifm_media = IFM_ETHER|IFM_NONE;
mii_mediachg(mii);
ifm->ifm_media = mtmp;
ifp->if_flags = itmp;
}
sc->bge_link = 0;
sc->bge_tx_saved_considx = BGE_TXCONS_UNSET;
ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
}
Reference in New Issue Copy Permalink