d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
5345bad07f
freebsd-skq/sys/dev/bge/if_bge.c

4169 lines
113 KiB
C
Raw Normal View History

Start each of the license/copyright comments with /*-, minor shuffle of lines
2005-01-06 01:43:34 +00:00
/*-
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* Copyright (c) 2001 Wind River Systems
* Copyright (c) 1997, 1998, 1999, 2001
* Bill Paul <wpaul@windriver.com>. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by Bill Paul.
* 4. Neither the name of the author nor the names of any co-contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
Use __FBSDID(). Also some minor style cleanups.
2003-08-24 17:55:58 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Broadcom BCM570x family gigabit ethernet driver for FreeBSD.
Use __FBSDID rather than rcsid[].
2003-04-03 21:36:33 +00:00
*
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* The Broadcom BCM5700 is based on technology originally developed by
* Alteon Networks as part of the Tigon I and Tigon II gigabit ethernet
* MAC chips. The BCM5700, sometimes refered to as the Tigon III, has
* two on-board MIPS R4000 CPUs and can have as much as 16MB of external
* SSRAM. The BCM5700 supports TCP, UDP and IP checksum offload, jumbo
* frames, highly configurable RX filtering, and 16 RX and TX queues
* (which, along with RX filter rules, can be used for QOS applications).
* Other features, such as TCP segmentation, may be available as part
* of value-added firmware updates. Unlike the Tigon I and Tigon II,
* firmware images can be stored in hardware and need not be compiled
* into the driver.
*
* The BCM5700 supports the PCI v2.2 and PCI-X v1.0 standards, and will
* function in a 32-bit/64-bit 33/66Mhz bus, or a 64-bit/133Mhz bus.
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
*
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* The BCM5701 is a single-chip solution incorporating both the BCM5700
Teach the bge driver about the BCM5701 (specifically, the 3Com 3c996B-T, with the 5701 rev B5 ASIC). One thing that confuses me still is that the 'link state change' bit in the status block seems to change state an awful lot. I have a workaround for this in place now, but it needs more investigation. For the moment though, this is enough to get the driver to work with this card.
2002-03-22 06:45:40 +00:00
* MAC and a BCM5401 10/100/1000 PHY. Unlike the BCM5700, the BCM5701
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* does not support external SSRAM.
*
* Broadcom also produces a variation of the BCM5700 under the "Altima"
* brand name, which is functionally similar but lacks PCI-X support.
*
* Without external SSRAM, you can only have at most 4 TX rings,
* and the use of the mini RX ring is disabled. This seems to imply
* that these features are simply not available on the BCM5701. As a
* result, this driver does not implement any support for the mini RX
* ring.
*/
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef HAVE_KERNEL_OPTION_HEADERS
#include "opt_device_polling.h"
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include <sys/param.h>
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
#include <sys/endian.h>
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include <sys/systm.h>
#include <sys/sockio.h>
#include <sys/mbuf.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
Add missing <sys/module.h> includes
2004-05-30 20:08:47 +00:00
#include <sys/module.h>
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include <sys/socket.h>
Allow the ASF feature to be disabled via a tunable. On one of my systems, bringing up the bge interface results in a complete system freeze when this feature is enabled. Leave it enabled by default.
2006-09-23 18:55:49 +00:00
#include <sys/sysctl.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 <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/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)
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
#define ETHER_MIN_NOPAD (ETHER_MIN_LEN - ETHER_CRC_LEN) /* i.e., 60 */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- Express hard dependencies on bus (pci, isa, pccard) and network layer (ether). - Don't abuse module names to facilitate ifconfig module loading; such abuse isn't really needed. (And if we do need type information associated with a module then we should make it explicit and not use hacks.)
2003-04-15 06:37:30 +00:00
MODULE_DEPEND(bge, pci, 1, 1, 1);
MODULE_DEPEND(bge, ether, 1, 1, 1);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
MODULE_DEPEND(bge, miibus, 1, 1, 1);
Replace FreeBSD 3.x syntax (controller miibus0) with 4.x syntax (device miibus) in time for 7.0 :-)
2005-10-22 05:06:55 +00:00
/* "device miibus" required. See GENERIC if you get errors here. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#include "miibus_if.h"
/*
* Various supported device vendors/types and their names. Note: the
* spec seems to indicate that the hardware still has Alteon's vendor
* ID burned into it, though it will always be overriden by the vendor
* ID in the EEPROM. Just to be safe, we cover all possibilities.
*/
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
static struct bge_type {
uint16_t bge_vid;
uint16_t bge_did;
} bge_devs[] = {
{ ALTEON_VENDORID, ALTEON_DEVICEID_BCM5700 },
{ ALTEON_VENDORID, ALTEON_DEVICEID_BCM5701 },
{ ALTIMA_VENDORID, ALTIMA_DEVICE_AC1000 },
{ ALTIMA_VENDORID, ALTIMA_DEVICE_AC1002 },
{ ALTIMA_VENDORID, ALTIMA_DEVICE_AC9100 },
{ APPLE_VENDORID, APPLE_DEVICE_BCM5701 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5700 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5701 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5702 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5702_ALT },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5702X },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5703 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5703_ALT },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5703X },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5704C },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5704S },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5704S_ALT },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705F },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705K },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705M },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5705M_ALT },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5714C },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5714S },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5715 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5715S },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5720 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5721 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5750 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5750M },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5751 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5751F },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5751M },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5752 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5752M },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5753 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5753F },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5753M },
- Recognize more device IDs adding support for BCM5754, BCM5755, BCM5787 based NICs. - Recognize BCM5703 B0 ASIC. - Rewrite the jumbo capability matching macro, so that chips known to work are listed there. [*] [*] I'm still not sure about this. Probably more corrections will be done to this macro after discussion with davidch@ and brad@OpenBSD. Obtained from: OpenBSD (brad)
2006-06-28 09:12:29 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5754 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5754M },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5755 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5755M },
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5780 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5780S },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5781 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5782 },
- Recognize more device IDs adding support for BCM5754, BCM5755, BCM5787 based NICs. - Recognize BCM5703 B0 ASIC. - Rewrite the jumbo capability matching macro, so that chips known to work are listed there. [*] [*] I'm still not sure about this. Probably more corrections will be done to this macro after discussion with davidch@ and brad@OpenBSD. Obtained from: OpenBSD (brad)
2006-06-28 09:12:29 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5786 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5787 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5787M },
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5788 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5789 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5901 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5901A2 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5903M },
{ SK_VENDORID, SK_DEVICEID_ALTIMA },
{ TC_VENDORID, TC_DEVICEID_3C996 },
{ 0, 0 }
};
static const struct bge_vendor {
uint16_t v_id;
const char *v_name;
} bge_vendors[] = {
{ ALTEON_VENDORID, "Alteon" },
{ ALTIMA_VENDORID, "Altima" },
{ APPLE_VENDORID, "Apple" },
{ BCOM_VENDORID, "Broadcom" },
{ SK_VENDORID, "SysKonnect" },
{ TC_VENDORID, "3Com" },
{ 0, NULL }
};
static const struct bge_revision {
uint32_t br_chipid;
const char *br_name;
} bge_revisions[] = {
{ BGE_CHIPID_BCM5700_A0, "BCM5700 A0" },
{ BGE_CHIPID_BCM5700_A1, "BCM5700 A1" },
{ BGE_CHIPID_BCM5700_B0, "BCM5700 B0" },
{ BGE_CHIPID_BCM5700_B1, "BCM5700 B1" },
{ BGE_CHIPID_BCM5700_B2, "BCM5700 B2" },
{ BGE_CHIPID_BCM5700_B3, "BCM5700 B3" },
{ BGE_CHIPID_BCM5700_ALTIMA, "BCM5700 Altima" },
{ BGE_CHIPID_BCM5700_C0, "BCM5700 C0" },
{ BGE_CHIPID_BCM5701_A0, "BCM5701 A0" },
{ BGE_CHIPID_BCM5701_B0, "BCM5701 B0" },
{ BGE_CHIPID_BCM5701_B2, "BCM5701 B2" },
{ BGE_CHIPID_BCM5701_B5, "BCM5701 B5" },
{ BGE_CHIPID_BCM5703_A0, "BCM5703 A0" },
{ BGE_CHIPID_BCM5703_A1, "BCM5703 A1" },
{ BGE_CHIPID_BCM5703_A2, "BCM5703 A2" },
{ BGE_CHIPID_BCM5703_A3, "BCM5703 A3" },
- Recognize more device IDs adding support for BCM5754, BCM5755, BCM5787 based NICs. - Recognize BCM5703 B0 ASIC. - Rewrite the jumbo capability matching macro, so that chips known to work are listed there. [*] [*] I'm still not sure about this. Probably more corrections will be done to this macro after discussion with davidch@ and brad@OpenBSD. Obtained from: OpenBSD (brad)
2006-06-28 09:12:29 +00:00
{ BGE_CHIPID_BCM5703_B0, "BCM5703 B0" },
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
{ BGE_CHIPID_BCM5704_A0, "BCM5704 A0" },
{ BGE_CHIPID_BCM5704_A1, "BCM5704 A1" },
{ BGE_CHIPID_BCM5704_A2, "BCM5704 A2" },
{ BGE_CHIPID_BCM5704_A3, "BCM5704 A3" },
{ BGE_CHIPID_BCM5704_B0, "BCM5704 B0" },
{ BGE_CHIPID_BCM5705_A0, "BCM5705 A0" },
{ BGE_CHIPID_BCM5705_A1, "BCM5705 A1" },
{ BGE_CHIPID_BCM5705_A2, "BCM5705 A2" },
{ BGE_CHIPID_BCM5705_A3, "BCM5705 A3" },
{ BGE_CHIPID_BCM5750_A0, "BCM5750 A0" },
{ BGE_CHIPID_BCM5750_A1, "BCM5750 A1" },
{ BGE_CHIPID_BCM5750_A3, "BCM5750 A3" },
{ BGE_CHIPID_BCM5750_B0, "BCM5750 B0" },
{ BGE_CHIPID_BCM5750_B1, "BCM5750 B1" },
{ BGE_CHIPID_BCM5750_C0, "BCM5750 C0" },
{ BGE_CHIPID_BCM5750_C1, "BCM5750 C1" },
Recognize the 5750 C2. Obtained from: OpenBSD (brad)
2006-08-07 12:51:50 +00:00
{ BGE_CHIPID_BCM5750_C2, "BCM5750 C2" },
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
{ BGE_CHIPID_BCM5714_A0, "BCM5714 A0" },
{ BGE_CHIPID_BCM5752_A0, "BCM5752 A0" },
{ BGE_CHIPID_BCM5752_A1, "BCM5752 A1" },
{ BGE_CHIPID_BCM5752_A2, "BCM5752 A2" },
{ BGE_CHIPID_BCM5714_B0, "BCM5714 B0" },
{ BGE_CHIPID_BCM5714_B3, "BCM5714 B3" },
{ BGE_CHIPID_BCM5715_A0, "BCM5715 A0" },
{ BGE_CHIPID_BCM5715_A1, "BCM5715 A1" },
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
/* 5784 and 5787 share the same ASIC ID */
{ BGE_CHIPID_BCM5787_A0, "BCM5754/5787 A0" },
{ BGE_CHIPID_BCM5787_A1, "BCM5754/5787 A1" },
{ BGE_CHIPID_BCM5787_A2, "BCM5754/5787 A2" },
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
{ 0, NULL }
};
/*
* Some defaults for major revisions, so that newer steppings
* that we don't know about have a shot at working.
*/
static const struct bge_revision bge_majorrevs[] = {
- Recognize more device IDs adding support for BCM5754, BCM5755, BCM5787 based NICs. - Recognize BCM5703 B0 ASIC. - Rewrite the jumbo capability matching macro, so that chips known to work are listed there. [*] [*] I'm still not sure about this. Probably more corrections will be done to this macro after discussion with davidch@ and brad@OpenBSD. Obtained from: OpenBSD (brad)
2006-06-28 09:12:29 +00:00
{ BGE_ASICREV_BCM5700, "unknown BCM5700" },
{ BGE_ASICREV_BCM5701, "unknown BCM5701" },
{ BGE_ASICREV_BCM5703, "unknown BCM5703" },
{ BGE_ASICREV_BCM5704, "unknown BCM5704" },
{ BGE_ASICREV_BCM5705, "unknown BCM5705" },
{ BGE_ASICREV_BCM5750, "unknown BCM5750" },
{ BGE_ASICREV_BCM5714_A0, "unknown BCM5714" },
{ BGE_ASICREV_BCM5752, "unknown BCM5752" },
{ BGE_ASICREV_BCM5780, "unknown BCM5780" },
{ BGE_ASICREV_BCM5714, "unknown BCM5714" },
{ BGE_ASICREV_BCM5755, "unknown BCM5755" },
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
/* 5784 and 5787 share the same ASIC ID */
{ BGE_ASICREV_BCM5787, "unknown BCM5754/5787" },
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
{ 0, 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
};
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
#define BGE_IS_JUMBO_CAPABLE(sc) ((sc)->bge_flags & BGE_FLAG_JUMBO)
#define BGE_IS_5700_FAMILY(sc) ((sc)->bge_flags & BGE_FLAG_5700_FAMILY)
#define BGE_IS_5705_PLUS(sc) ((sc)->bge_flags & BGE_FLAG_5705_PLUS)
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
#define BGE_IS_5714_FAMILY(sc) ((sc)->bge_flags & BGE_FLAG_5714_FAMILY)
#define BGE_IS_575X_PLUS(sc) ((sc)->bge_flags & BGE_FLAG_575X_PLUS)
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
const struct bge_revision * bge_lookup_rev(uint32_t);
const struct bge_vendor * bge_lookup_vendor(uint16_t);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static int bge_probe(device_t);
static int bge_attach(device_t);
static int bge_detach(device_t);
static int bge_suspend(device_t);
static int bge_resume(device_t);
static void bge_release_resources(struct bge_softc *);
static void bge_dma_map_addr(void *, bus_dma_segment_t *, int, int);
static int bge_dma_alloc(device_t);
static void bge_dma_free(struct bge_softc *);
static void bge_txeof(struct bge_softc *);
static void bge_rxeof(struct bge_softc *);
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
static void bge_asf_driver_up (struct bge_softc *);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static void bge_tick(void *);
static void bge_stats_update(struct bge_softc *);
static void bge_stats_update_regs(struct bge_softc *);
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
static int bge_encap(struct bge_softc *, struct mbuf **, uint32_t *);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static void bge_intr(void *);
static void bge_start_locked(struct ifnet *);
static void bge_start(struct ifnet *);
static int bge_ioctl(struct ifnet *, u_long, caddr_t);
static void bge_init_locked(struct bge_softc *);
static void bge_init(void *);
static void bge_stop(struct bge_softc *);
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
static void bge_watchdog(struct bge_softc *);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static void bge_shutdown(device_t);
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +00:00
static int bge_ifmedia_upd_locked(struct ifnet *);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static int bge_ifmedia_upd(struct ifnet *);
static void bge_ifmedia_sts(struct ifnet *, struct ifmediareq *);
static uint8_t bge_eeprom_getbyte(struct bge_softc *, int, uint8_t *);
static int bge_read_eeprom(struct bge_softc *, caddr_t, int, int);
Do not strip VLAN tag in promiscuous mode.
2006-09-18 22:18:21 +00:00
static void bge_setpromisc(struct bge_softc *);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static void bge_setmulti(struct bge_softc *);
static int bge_newbuf_std(struct bge_softc *, int, struct mbuf *);
static int bge_newbuf_jumbo(struct bge_softc *, int, struct mbuf *);
static int bge_init_rx_ring_std(struct bge_softc *);
static void bge_free_rx_ring_std(struct bge_softc *);
static int bge_init_rx_ring_jumbo(struct bge_softc *);
static void bge_free_rx_ring_jumbo(struct bge_softc *);
static void bge_free_tx_ring(struct bge_softc *);
static int bge_init_tx_ring(struct bge_softc *);
static int bge_chipinit(struct bge_softc *);
static int bge_blockinit(struct bge_softc *);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static uint32_t bge_readmem_ind(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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static uint32_t bge_readreg_ind(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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static void bge_writereg_ind(struct bge_softc *, int, int);
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
static void bge_writemem_direct(struct bge_softc *, int, int) __unused;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static int bge_miibus_readreg(device_t, int, int);
static int bge_miibus_writereg(device_t, int, int, int);
static void bge_miibus_statchg(device_t);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static void bge_poll(struct ifnet *ifp, enum poll_cmd cmd, int count);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +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
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
#define BGE_RESET_START 1
#define BGE_RESET_STOP 2
static void bge_sig_post_reset(struct bge_softc *, int);
static void bge_sig_legacy(struct bge_softc *, int);
static void bge_sig_pre_reset(struct bge_softc *, int);
static int bge_reset(struct bge_softc *);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static void bge_link_upd(struct bge_softc *);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
/*
* The BGE_REGISTER_DEBUG option is only for low-level debugging. It may
* leak information to untrusted users. It is also known to cause alignment
* traps on certain architectures.
*/
#ifdef BGE_REGISTER_DEBUG
static int bge_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
static int bge_sysctl_reg_read(SYSCTL_HANDLER_ARGS);
static int bge_sysctl_mem_read(SYSCTL_HANDLER_ARGS);
#endif
static void bge_add_sysctls(struct bge_softc *);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static device_method_t bge_methods[] = {
/* Device interface */
DEVMETHOD(device_probe, bge_probe),
DEVMETHOD(device_attach, bge_attach),
DEVMETHOD(device_detach, bge_detach),
DEVMETHOD(device_shutdown, bge_shutdown),
Implement suspend/resume methods to be more ACPI friendly. I'm able to suspend/resume my laptop without this change, but then I need to wait for the watchdog to reset the card. With this change, it is ready immediately. Glanced at by: glebius
2005-09-28 19:20:49 +00:00
DEVMETHOD(device_suspend, bge_suspend),
DEVMETHOD(device_resume, bge_resume),
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* bus interface */
DEVMETHOD(bus_print_child, bus_generic_print_child),
DEVMETHOD(bus_driver_added, bus_generic_driver_added),
/* MII interface */
DEVMETHOD(miibus_readreg, bge_miibus_readreg),
DEVMETHOD(miibus_writereg, bge_miibus_writereg),
DEVMETHOD(miibus_statchg, bge_miibus_statchg),
{ 0, 0 }
};
static driver_t bge_driver = {
"bge",
bge_methods,
sizeof(struct bge_softc)
};
static devclass_t bge_devclass;
- Express hard dependencies on bus (pci, isa, pccard) and network layer (ether). - Don't abuse module names to facilitate ifconfig module loading; such abuse isn't really needed. (And if we do need type information associated with a module then we should make it explicit and not use hacks.)
2003-04-15 06:37:30 +00:00
DRIVER_MODULE(bge, pci, bge_driver, bge_devclass, 0, 0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
DRIVER_MODULE(miibus, bge, miibus_driver, miibus_devclass, 0, 0);
make BGE_FAKE_AUTONEG a tunable. This allows one to change the behavior of the driver pre-boot. NOTE: This patch was made for DragonFly BSD by Sepherosa Ziehau. PR: kern/94833 Submitted by: Devon H. O'Dell Obtained from: DragonFly MFC after: 1 month
2006-04-25 15:56:52 +00:00
static int bge_fake_autoneg = 0;
Allow the ASF feature to be disabled via a tunable. On one of my systems, bringing up the bge interface results in a complete system freeze when this feature is enabled. Leave it enabled by default.
2006-09-23 18:55:49 +00:00
static int bge_allow_asf = 1;
make BGE_FAKE_AUTONEG a tunable. This allows one to change the behavior of the driver pre-boot. NOTE: This patch was made for DragonFly BSD by Sepherosa Ziehau. PR: kern/94833 Submitted by: Devon H. O'Dell Obtained from: DragonFly MFC after: 1 month
2006-04-25 15:56:52 +00:00
TUNABLE_INT("hw.bge.fake_autoneg", &bge_fake_autoneg);
Allow the ASF feature to be disabled via a tunable. On one of my systems, bringing up the bge interface results in a complete system freeze when this feature is enabled. Leave it enabled by default.
2006-09-23 18:55:49 +00:00
TUNABLE_INT("hw.bge.allow_asf", &bge_allow_asf);
SYSCTL_NODE(_hw, OID_AUTO, bge, CTLFLAG_RD, 0, "BGE driver parameters");
SYSCTL_INT(_hw_bge, OID_AUTO, fake_autoneg, CTLFLAG_RD, &bge_fake_autoneg, 0,
"Enable fake autonegotiation for certain blade systems");
SYSCTL_INT(_hw_bge, OID_AUTO, allow_asf, CTLFLAG_RD, &bge_allow_asf, 0,
"Allow ASF mode if available");
make BGE_FAKE_AUTONEG a tunable. This allows one to change the behavior of the driver pre-boot. NOTE: This patch was made for DragonFly BSD by Sepherosa Ziehau. PR: kern/94833 Submitted by: Devon H. O'Dell Obtained from: DragonFly MFC after: 1 month
2006-04-25 15:56:52 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static uint32_t
bge_readmem_ind(struct bge_softc *sc, int off)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
device_t dev;
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
uint32_t val;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
dev = sc->bge_dev;
pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, off, 4);
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
val = pci_read_config(dev, BGE_PCI_MEMWIN_DATA, 4);
pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, 0, 4);
return (val);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_writemem_ind(struct bge_softc *sc, int off, int val)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
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);
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
pci_write_config(dev, BGE_PCI_MEMWIN_BASEADDR, 0, 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 notdef
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static uint32_t
bge_readreg_ind(struct bge_softc *sc, int off)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
device_t dev;
dev = sc->bge_dev;
pci_write_config(dev, BGE_PCI_REG_BASEADDR, off, 4);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (pci_read_config(dev, BGE_PCI_REG_DATA, 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
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_writereg_ind(struct bge_softc *sc, int off, int val)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
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);
}
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
static void
bge_writemem_direct(struct bge_softc *sc, int off, int val)
{
CSR_WRITE_4(sc, off, val);
}
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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
{
struct bge_dmamap_arg *ctx;
if (error)
return;
ctx = arg;
if (nseg > ctx->bge_maxsegs) {
ctx->bge_maxsegs = 0;
return;
}
ctx->bge_busaddr = segs->ds_addr;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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.
*/
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
static uint8_t
bge_eeprom_getbyte(struct bge_softc *sc, int addr, uint8_t *dest)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t byte = 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 use of auto EEPROM access so we can avoid
* having to use the bitbang method.
*/
BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_AUTO_EEPROM);
/* Reset the EEPROM, load the clock period. */
CSR_WRITE_4(sc, BGE_EE_ADDR,
BGE_EEADDR_RESET|BGE_EEHALFCLK(BGE_HALFCLK_384SCL));
DELAY(20);
/* Issue the read EEPROM command. */
CSR_WRITE_4(sc, BGE_EE_ADDR, BGE_EE_READCMD | addr);
/* Wait for completion */
for(i = 0; i < BGE_TIMEOUT * 10; i++) {
DELAY(10);
if (CSR_READ_4(sc, BGE_EE_ADDR) & BGE_EEADDR_DONE)
break;
}
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "EEPROM read timed out\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (1);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* Get result. */
byte = CSR_READ_4(sc, BGE_EE_DATA);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
*dest = (byte >> ((addr % 4) * 8)) & 0xFF;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* Read a sequence of bytes from the EEPROM.
*/
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_read_eeprom(struct bge_softc *sc, caddr_t dest, int off, int 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
{
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
int i, error = 0;
uint8_t byte = 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
for (i = 0; i < cnt; i++) {
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
error = bge_eeprom_getbyte(sc, off + i, &byte);
if (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
break;
*(dest + i) = byte;
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (error ? 1 : 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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_miibus_readreg(device_t dev, int phy, int reg)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_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)
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +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) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->bge_dev, "PHY read timed out\n");
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
val = 0;
goto done;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
val = CSR_READ_4(sc, BGE_MI_COMM);
- Move bge_phy_hack into the phy code and implement the various DSP patch workarounds for each phy revision. Obtained from: NetBSD & Broadcom Linux driver - Disable AUTOPOLL when accessing the PHY as it may cause PCI errors. Obtained from: NetBSD - Check the UPDATED bit in the status block so the driver knows that the status block as indeed changed since the last access. Broadcom documentation states drivers should unset the UPDATED/CHANGED bits after reading them. - When changing media types, first loop the phy then set the media. Broadcom documentation and Linux drivers do this and I observed much better handling of link after this change. - Broadcom documentation states that for 1000BaseT operation, autonegotiation must be enabled. Fix hard coding of media so that the driver only advertises 1000BaseT as the supported media type and enable autonegotition. - Only set Master/Slave on the 5701. Obtained from Broadcom Linux driver.
2003-05-03 19:06:50 +00:00
done:
if (autopoll & BGE_MIMODE_AUTOPOLL) {
BGE_SETBIT(sc, BGE_MI_MODE, BGE_MIMODE_AUTOPOLL);
DELAY(40);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (val & BGE_MICOMM_READFAIL)
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (val & 0xFFFF);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_miibus_writereg(device_t dev, int phy, int reg, int val)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_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) {
- Consistently use if_printf() only in interface methods: if_start(), if_watchdog, etc., or in functions used only in these methods. In all other functions in the driver use device_printf(). - Use __func__ instead of typing function name. Submitted by: Alex Lyashkov <umka sevcity.net>
2006-09-15 15:16:12 +00:00
device_printf(sc->bge_dev, "PHY read timed out\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_miibus_statchg(device_t dev)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
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);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +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);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +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_MAC_MODE, BGE_PORTMODE_MII);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
if ((mii->mii_media_active & IFM_GMASK) == IFM_FDX)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_MAC_MODE, BGE_MACMODE_HALF_DUPLEX);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +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_MAC_MODE, BGE_MACMODE_HALF_DUPLEX);
}
/*
* Intialize a standard receive ring descriptor.
*/
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_newbuf_std(struct bge_softc *sc, int i, struct mbuf *m)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
struct mbuf *m_new = NULL;
struct bge_rx_bd *r;
struct bge_dmamap_arg ctx;
int error;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (m == NULL) {
Use m_getcl() instead of getting mbuf, and then getting a cluster.
2006-01-23 15:57:02 +00:00
m_new = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
if (m_new == NULL)
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
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
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;
}
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if ((sc->bge_flags & BGE_FLAG_RX_ALIGNBUG) == 0)
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
m_adj(m_new, ETHER_ALIGN);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc->bge_cdata.bge_rx_std_chain[i] = m_new;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
r = &sc->bge_ldata.bge_rx_std_ring[i];
ctx.bge_maxsegs = 1;
ctx.sc = sc;
error = bus_dmamap_load(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i], mtod(m_new, void *),
m_new->m_len, bge_dma_map_addr, &ctx, BUS_DMA_NOWAIT);
if (error || ctx.bge_maxsegs == 0) {
If bus_dmamap_load() failed, we free the mbuf. We also need to clear the pointer, to avoid double free on next bge_stop().
2005-11-30 12:37:07 +00:00
if (m == NULL) {
sc->bge_cdata.bge_rx_std_chain[i] = NULL;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
m_freem(m_new);
If bus_dmamap_load() failed, we free the mbuf. We also need to clear the pointer, to avoid double free on next bge_stop().
2005-11-30 12:37:07 +00:00
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
}
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
r->bge_addr.bge_addr_lo = BGE_ADDR_LO(ctx.bge_busaddr);
r->bge_addr.bge_addr_hi = BGE_ADDR_HI(ctx.bge_busaddr);
r->bge_flags = BGE_RXBDFLAG_END;
r->bge_len = m_new->m_len;
r->bge_idx = i;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i],
BUS_DMASYNC_PREREAD);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* Initialize a jumbo receive ring descriptor. This allocates
* a jumbo buffer from the pool managed internally by the driver.
*/
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_newbuf_jumbo(struct bge_softc *sc, int i, struct mbuf *m)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
bus_dma_segment_t segs[BGE_NSEG_JUMBO];
struct bge_extrx_bd *r;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
struct mbuf *m_new = NULL;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
int nsegs;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
int error;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (m == NULL) {
Back out M_* changes, per decision of the TRB. Approved by: trb
2003-02-19 05:47:46 +00:00
MGETHDR(m_new, M_DONTWAIT, MT_DATA);
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
if (m_new == NULL)
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
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
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
m_cljget(m_new, M_DONTWAIT, MJUM9BYTES);
if (!(m_new->m_flags & M_EXT)) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m_freem(m_new);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
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
}
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
m_new->m_len = m_new->m_pkthdr.len = MJUM9BYTES;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
} else {
m_new = m;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
m_new->m_len = m_new->m_pkthdr.len = MJUM9BYTES;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m_new->m_data = m_new->m_ext.ext_buf;
}
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if ((sc->bge_flags & BGE_FLAG_RX_ALIGNBUG) == 0)
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
m_adj(m_new, ETHER_ALIGN);
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i],
m_new, segs, &nsegs, BUS_DMA_NOWAIT);
if (error) {
if (m == NULL)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
m_freem(m_new);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (error);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
}
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
sc->bge_cdata.bge_rx_jumbo_chain[i] = m_new;
/*
* Fill in the extended RX buffer descriptor.
*/
r = &sc->bge_ldata.bge_rx_jumbo_ring[i];
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
r->bge_flags = BGE_RXBDFLAG_JUMBO_RING|BGE_RXBDFLAG_END;
r->bge_idx = i;
Be ready to a case when not a constant number of segments is returned by bus_dmamap_load_mbuf_sg() on jumbo buffer allocation. Reviewed by: scottl, gallatin
2006-01-18 14:31:21 +00:00
r->bge_len3 = r->bge_len2 = r->bge_len1 = 0;
switch (nsegs) {
case 4:
r->bge_addr3.bge_addr_lo = BGE_ADDR_LO(segs[3].ds_addr);
r->bge_addr3.bge_addr_hi = BGE_ADDR_HI(segs[3].ds_addr);
r->bge_len3 = segs[3].ds_len;
case 3:
r->bge_addr2.bge_addr_lo = BGE_ADDR_LO(segs[2].ds_addr);
r->bge_addr2.bge_addr_hi = BGE_ADDR_HI(segs[2].ds_addr);
r->bge_len2 = segs[2].ds_len;
case 2:
r->bge_addr1.bge_addr_lo = BGE_ADDR_LO(segs[1].ds_addr);
r->bge_addr1.bge_addr_hi = BGE_ADDR_HI(segs[1].ds_addr);
r->bge_len1 = segs[1].ds_len;
case 1:
r->bge_addr0.bge_addr_lo = BGE_ADDR_LO(segs[0].ds_addr);
r->bge_addr0.bge_addr_hi = BGE_ADDR_HI(segs[0].ds_addr);
r->bge_len0 = segs[0].ds_len;
break;
default:
panic("%s: %d segments\n", __func__, nsegs);
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_jumbo_dmamap[i],
BUS_DMASYNC_PREREAD);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* The standard receive ring has 512 entries in it. At 2K per mbuf cluster,
* that's 1MB or memory, which is a lot. For now, we fill only the first
* 256 ring entries and hope that our CPU is fast enough to keep up with
* the NIC.
*/
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_init_rx_ring_std(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
{
int i;
for (i = 0; i < BGE_SSLOTS; i++) {
if (bge_newbuf_std(sc, i, NULL) == ENOBUFS)
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
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
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);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_free_rx_ring_std(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
{
int i;
for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
if (sc->bge_cdata.bge_rx_std_chain[i] != NULL) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i],
BUS_DMASYNC_POSTREAD);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i]);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
m_freem(sc->bge_cdata.bge_rx_std_chain[i]);
sc->bge_cdata.bge_rx_std_chain[i] = NULL;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bzero((char *)&sc->bge_ldata.bge_rx_std_ring[i],
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sizeof(struct bge_rx_bd));
}
}
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_init_rx_ring_jumbo(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 bge_rcb *rcb;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
int i;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
if (bge_newbuf_jumbo(sc, i, NULL) == ENOBUFS)
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
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
bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map,
BUS_DMASYNC_PREREAD|BUS_DMASYNC_PREWRITE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc->bge_jumbo = i - 1;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
rcb = &sc->bge_ldata.bge_info.bge_jumbo_rx_rcb;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(0,
BGE_RCB_FLAG_USE_EXT_RX_BD);
Don't reference the maxlen and flags fields in the ring control blocks as separate 16-bit entities. Some of the ring control blocks are in NIC memory, so they must be referenced using 32-bit accesses. Smaller accesses have been observed to fail under some conditions. This caused the rings to be set up wrong, leading to writes by the card outside of the intended bounds of the rings. This problem was diagnosed by Michael Barthelow. Don Bowman submitted a patch which fixed the problem using a slightly different approach. Reference ring control blocks in NIC memory using a pointer to volatile. Parenthesize the BGE_HOSTADDR macro definition properly. MFC after: 3 days
2003-01-06 23:46:47 +00:00
CSR_WRITE_4(sc, BGE_RX_JUMBO_RCB_MAXLEN_FLAGS, rcb->bge_maxlen_flags);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MBX_RX_JUMBO_PROD_LO, sc->bge_jumbo);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_free_rx_ring_jumbo(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
{
int i;
for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
if (sc->bge_cdata.bge_rx_jumbo_chain[i] != NULL) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i],
BUS_DMASYNC_POSTREAD);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
m_freem(sc->bge_cdata.bge_rx_jumbo_chain[i]);
sc->bge_cdata.bge_rx_jumbo_chain[i] = NULL;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bzero((char *)&sc->bge_ldata.bge_rx_jumbo_ring[i],
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
sizeof(struct bge_extrx_bd));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
}
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_free_tx_ring(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
{
int i;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
if (sc->bge_ldata.bge_tx_ring == NULL)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
for (i = 0; i < BGE_TX_RING_CNT; i++) {
if (sc->bge_cdata.bge_tx_chain[i] != NULL) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[i],
BUS_DMASYNC_POSTWRITE);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[i]);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
m_freem(sc->bge_cdata.bge_tx_chain[i]);
sc->bge_cdata.bge_tx_chain[i] = NULL;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bzero((char *)&sc->bge_ldata.bge_tx_ring[i],
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sizeof(struct bge_tx_bd));
}
}
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_init_tx_ring(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
{
sc->bge_txcnt = 0;
sc->bge_tx_saved_considx = 0;
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
/* Initialize transmit producer index for host-memory send ring. */
sc->bge_tx_prodidx = 0;
CSR_WRITE_4(sc, BGE_MBX_TX_HOST_PROD0_LO, sc->bge_tx_prodidx);
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
/* 5700 b2 errata */
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
CSR_WRITE_4(sc, BGE_MBX_TX_HOST_PROD0_LO, sc->bge_tx_prodidx);
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
/* NIC-memory send ring not used; initialize to zero. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MBX_TX_NIC_PROD0_LO, 0);
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
/* 5700 b2 errata */
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
CSR_WRITE_4(sc, BGE_MBX_TX_NIC_PROD0_LO, 0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Do not strip VLAN tag in promiscuous mode.
2006-09-18 22:18:21 +00:00
static void
bge_setpromisc(struct bge_softc *sc)
{
struct ifnet *ifp;
BGE_LOCK_ASSERT(sc);
ifp = sc->bge_ifp;
/*
* Enable or disable promiscuous mode as needed.
* Do not strip VLAN tag when promiscuous mode is enabled.
*/
if (ifp->if_flags & IFF_PROMISC)
BGE_SETBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_PROMISC |
BGE_RXMODE_RX_KEEP_VLAN_DIAG);
else
BGE_CLRBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_PROMISC |
BGE_RXMODE_RX_KEEP_VLAN_DIAG);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_setmulti(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;
struct ifmultiaddr *ifma;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t hashes[4] = { 0, 0, 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
int h, i;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (ifp->if_flags & IFF_ALLMULTI || ifp->if_flags & IFF_PROMISC) {
for (i = 0; i < 4; i++)
CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), 0xFFFFFFFF);
return;
}
/* First, zot all the existing filters. */
for (i = 0; i < 4; i++)
CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), 0);
/* Now program new ones. */
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
IF_ADDR_LOCK(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
if (ifma->ifma_addr->sa_family != AF_LINK)
continue;
Replace handrolled CRC calculation with ether_crc32_[lb]e().
2004-06-09 14:34:04 +00:00
h = ether_crc32_le(LLADDR((struct sockaddr_dl *)
ifma->ifma_addr), ETHER_ADDR_LEN) & 0x7F;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
hashes[(h & 0x60) >> 5] |= 1 << (h & 0x1F);
}
Modify device drivers supporting multicast addresses to lock if_addr_mtx over iteration of their multicast address lists when synchronizing the hardware address filter with the network stack-maintained list. Problem reported by: Ed Maste (emaste at phaedrus dot sandvine dot ca> MFC after: 1 week
2005-08-03 00:18:35 +00:00
IF_ADDR_UNLOCK(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
for (i = 0; i < 4; i++)
CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), hashes[i]);
}
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
static void
bge_sig_pre_reset(sc, type)
struct bge_softc *sc;
int type;
{
/*
* Some chips don't like this so only do this if ASF is enabled
*/
if (sc->bge_asf_mode)
bge_writemem_ind(sc, BGE_SOFTWARE_GENCOMM, BGE_MAGIC_NUMBER);
if (sc->bge_asf_mode & ASF_NEW_HANDSHAKE) {
switch (type) {
case BGE_RESET_START:
bge_writemem_ind(sc, BGE_SDI_STATUS, 0x1); /* START */
break;
case BGE_RESET_STOP:
bge_writemem_ind(sc, BGE_SDI_STATUS, 0x2); /* UNLOAD */
break;
}
}
}
static void
bge_sig_post_reset(sc, type)
struct bge_softc *sc;
int type;
{
if (sc->bge_asf_mode & ASF_NEW_HANDSHAKE) {
switch (type) {
case BGE_RESET_START:
bge_writemem_ind(sc, BGE_SDI_STATUS, 0x80000001);
/* START DONE */
break;
case BGE_RESET_STOP:
bge_writemem_ind(sc, BGE_SDI_STATUS, 0x80000002);
break;
}
}
}
static void
bge_sig_legacy(sc, type)
struct bge_softc *sc;
int type;
{
if (sc->bge_asf_mode) {
switch (type) {
case BGE_RESET_START:
bge_writemem_ind(sc, BGE_SDI_STATUS, 0x1); /* START */
break;
case BGE_RESET_STOP:
bge_writemem_ind(sc, BGE_SDI_STATUS, 0x2); /* UNLOAD */
break;
}
}
}
void bge_stop_fw(struct bge_softc *);
void
bge_stop_fw(sc)
struct bge_softc *sc;
{
int i;
if (sc->bge_asf_mode) {
bge_writemem_ind(sc, BGE_SOFTWARE_GENCOMM_FW, BGE_FW_PAUSE);
CSR_WRITE_4(sc, BGE_CPU_EVENT,
CSR_READ_4(sc, BGE_CPU_EVENT) != (1 << 14));
for (i = 0; i < 100; i++ ) {
if (!(CSR_READ_4(sc, BGE_CPU_EVENT) & (1 << 14)))
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
/*
* Do endian, PCI and DMA initialization. Also check the on-board ROM
* self-test results.
*/
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_chipinit(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
{
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t dma_rw_ctl;
int i;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
/* Set endianness before we access any non-PCI registers. */
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
pci_write_config(sc->bge_dev, BGE_PCI_MISC_CTL, BGE_INIT, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Check the 'ROM failed' bit on the RX CPU to see if
* self-tests passed.
*/
if (CSR_READ_4(sc, BGE_RXCPU_MODE) & BGE_RXCPUMODE_ROMFAIL) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "RX CPU self-diagnostics failed!\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (ENODEV);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* 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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
i < BGE_STATS_BLOCK_END + 1; i += sizeof(uint32_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_MEMWIN_WRITE(sc, i, 0);
for (i = BGE_STATUS_BLOCK;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
i < BGE_STATUS_BLOCK_END + 1; i += sizeof(uint32_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_MEMWIN_WRITE(sc, i, 0);
/* Set up the PCI DMA control register. */
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_PCIE) {
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
/* PCI Express bus */
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
dma_rw_ctl = BGE_PCI_READ_CMD|BGE_PCI_WRITE_CMD |
(0xf << BGE_PCIDMARWCTL_RD_WAT_SHIFT) |
(0x2 << BGE_PCIDMARWCTL_WR_WAT_SHIFT);
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
} else if (sc->bge_flags & BGE_FLAG_PCIX) {
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
/* PCI-X bus */
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
if (BGE_IS_5714_FAMILY(sc)) {
dma_rw_ctl = BGE_PCI_READ_CMD|BGE_PCI_WRITE_CMD;
dma_rw_ctl &= ~BGE_PCIDMARWCTL_ONEDMA_ATONCE; /* XXX */
/* XXX magic values, Broadcom-supplied Linux driver */
if (sc->bge_asicrev == BGE_ASICREV_BCM5780)
dma_rw_ctl |= (1 << 20) | (1 << 18) |
BGE_PCIDMARWCTL_ONEDMA_ATONCE;
else
dma_rw_ctl |= (1 << 20) | (1 << 18) | (1 << 15);
} else if (sc->bge_asicrev == BGE_ASICREV_BCM5704)
/*
* The 5704 uses a different encoding of read/write
* watermarks.
*/
- 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) {
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t tmp;
- 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
tmp = CSR_READ_4(sc, BGE_PCI_CLKCTL) & 0x1f;
if (tmp == 0x6 || tmp == 0x7)
dma_rw_ctl |= BGE_PCIDMARWCTL_ONEDMA_ATONCE;
}
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
} else
/* Conventional PCI bus */
dma_rw_ctl = BGE_PCI_READ_CMD|BGE_PCI_WRITE_CMD |
(0x7 << BGE_PCIDMARWCTL_RD_WAT_SHIFT) |
(0x7 << BGE_PCIDMARWCTL_WR_WAT_SHIFT) |
(0x0F);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 ||
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5705)
- Change the short hand representation of the various ASIC revisions - Implement the ONEDMA_AT_ONCE workaround as described in the 5703/5704 eratta documents. Obtained from: NetBSD & Broadcom documentation
2003-05-03 22:58:45 +00:00
dma_rw_ctl &= ~BGE_PCIDMARWCTL_MINDMA;
pci_write_config(sc->bge_dev, BGE_PCI_DMA_RW_CTL, dma_rw_ctl, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Set up general mode register.
*/
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
CSR_WRITE_4(sc, BGE_MODE_CTL, BGE_DMA_SWAP_OPTIONS|
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_MODECTL_MAC_ATTN_INTR|BGE_MODECTL_HOST_SEND_BDS|
Enable 'complete' rx checksum offloading (i.e. let chip calculate checksums with pseudo header for tcp/udp packets). This could save one in_pseudo() call per incoming tcp/udp packet. Approved by: glebius (mentor) MFC after: 3 weeks
2006-02-02 09:58:31 +00:00
BGE_MODECTL_TX_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
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
/*
* Tell the firmware the driver is running
*/
if (sc->bge_asf_mode & ASF_STACKUP)
BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
Follow the lead of recent changes in the 3com Linux driver by disabling memory write invalidate unconditionally. It looks like they've decided that MWI just doesn't work with these devices. Also, remove now-irrelevant code that set PCI write boundary values based on the cache line size. MFC after: 2 weeks
2002-09-22 18:58:58 +00:00
* Disable memory write invalidate. Apparently it is not supported
* properly by these devices.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
Follow the lead of recent changes in the 3com Linux driver by disabling memory write invalidate unconditionally. It looks like they've decided that MWI just doesn't work with these devices. Also, remove now-irrelevant code that set PCI write boundary values based on the cache line size. MFC after: 2 weeks
2002-09-22 18:58:58 +00:00
PCI_CLRBIT(sc->bge_dev, BGE_PCI_CMD, PCIM_CMD_MWIEN, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#ifdef __brokenalpha__
/*
* Must insure that we do not cross an 8K (bytes) boundary
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
* for DMA reads. Our highest limit is 1K bytes. This is a
* restriction on some ALPHA platforms with early revision
* 21174 PCI chipsets, such as the AlphaPC 164lx
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
Fix a couple of longstanding typos which would have made a statement fail to compile if it weren't inside "#ifdef __brokenalpha__". Submitted by: Birger Toedtmann <birger@takatukaland.de> MFC after: 2 weeks
2002-09-22 19:27:07 +00:00
PCI_SETBIT(sc->bge_dev, BGE_PCI_DMA_RW_CTL,
BGE_PCI_READ_BNDRY_1024BYTES, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
#endif
/* Set the timer prescaler (always 66Mhz) */
CSR_WRITE_4(sc, BGE_MISC_CFG, 65 << 1/*BGE_32BITTIME_66MHZ*/);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_blockinit(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 bge_rcb *rcb;
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
bus_size_t vrcb;
bge_hostaddr taddr;
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
uint32_t val;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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. */
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
/* Configure mbuf memory pool */
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_BASEADDR, BGE_BUFFPOOL_1);
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 device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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 */
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
CSR_WRITE_4(sc, BGE_BMAN_MODE,
BGE_BMANMODE_ENABLE|BGE_BMANMODE_LOMBUF_ATTN);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
/* Poll for buffer manager start indication */
for (i = 0; i < BGE_TIMEOUT; i++) {
if (CSR_READ_4(sc, BGE_BMAN_MODE) & BGE_BMANMODE_ENABLE)
break;
DELAY(10);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"buffer manager failed to start\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (ENXIO);
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
}
/* Enable flow-through queues */
CSR_WRITE_4(sc, BGE_FTQ_RESET, 0xFFFFFFFF);
CSR_WRITE_4(sc, BGE_FTQ_RESET, 0);
/* Wait until queue initialization is complete */
for (i = 0; i < BGE_TIMEOUT; i++) {
if (CSR_READ_4(sc, BGE_FTQ_RESET) == 0)
break;
DELAY(10);
}
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "flow-through queue init failed\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (ENXIO);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* 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 BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (BGE_IS_5705_PLUS(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
rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(512, 0);
else
rcb->bge_maxlen_flags =
BGE_RCB_MAXLEN_FLAGS(BGE_MAX_FRAMELEN, 0);
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
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).
*/
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
if (BGE_IS_JUMBO_CAPABLE(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
rcb = &sc->bge_ldata.bge_info.bge_jumbo_rx_rcb;
rcb->bge_hostaddr.bge_addr_lo =
BGE_ADDR_LO(sc->bge_ldata.bge_rx_jumbo_ring_paddr);
rcb->bge_hostaddr.bge_addr_hi =
BGE_ADDR_HI(sc->bge_ldata.bge_rx_jumbo_ring_paddr);
bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map,
BUS_DMASYNC_PREREAD);
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
rcb->bge_maxlen_flags = BGE_RCB_MAXLEN_FLAGS(0,
BGE_RCB_FLAG_USE_EXT_RX_BD|BGE_RCB_FLAG_RING_DISABLED);
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
rcb->bge_nicaddr = BGE_JUMBO_RX_RINGS;
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_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.
*/
Use the BGE_IS_* macros consistently. Also add a couple of missing cases to the debug_info function.
2006-12-13 21:03:55 +00:00
if (BGE_IS_5705_PLUS(sc))
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
val = 8;
else
val = BGE_STD_RX_RING_CNT / 8;
CSR_WRITE_4(sc, BGE_RBDI_STD_REPL_THRESH, val);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_RBDI_JUMBO_REPL_THRESH, BGE_JUMBO_RX_RING_CNT/8);
/*
* Disable all unused send rings by setting the 'ring disabled'
* bit in the flags field of all the TX send ring control blocks.
* These are located in NIC memory.
*/
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb = BGE_MEMWIN_START + BGE_SEND_RING_RCB;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
for (i = 0; i < BGE_TX_RINGS_EXTSSRAM_MAX; i++) {
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
BGE_RCB_MAXLEN_FLAGS(0, BGE_RCB_FLAG_RING_DISABLED));
RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0);
vrcb += sizeof(struct bge_rcb);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* Configure TX RCB 0 (we use only the first ring) */
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb = BGE_MEMWIN_START + BGE_SEND_RING_RCB;
BGE_HOSTADDR(taddr, sc->bge_ldata.bge_tx_ring_paddr);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, taddr.bge_addr_hi);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, taddr.bge_addr_lo);
RCB_WRITE_4(sc, vrcb, bge_nicaddr,
BGE_NIC_TXRING_ADDR(0, BGE_TX_RING_CNT));
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(sc)))
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
BGE_RCB_MAXLEN_FLAGS(BGE_TX_RING_CNT, 0));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Disable all unused RX return rings */
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb = BGE_MEMWIN_START + BGE_RX_RETURN_RING_RCB;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
for (i = 0; i < BGE_RX_RINGS_MAX; i++) {
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, 0);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, 0);
RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_RCB_MAXLEN_FLAGS(sc->bge_return_ring_cnt,
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
BGE_RCB_FLAG_RING_DISABLED));
RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MBX_RX_CONS0_LO +
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
(i * (sizeof(uint64_t))), 0);
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb += sizeof(struct bge_rcb);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* Initialize RX ring indexes */
CSR_WRITE_4(sc, BGE_MBX_RX_STD_PROD_LO, 0);
CSR_WRITE_4(sc, BGE_MBX_RX_JUMBO_PROD_LO, 0);
CSR_WRITE_4(sc, BGE_MBX_RX_MINI_PROD_LO, 0);
/*
* Set up RX return ring 0
* Note that the NIC address for RX return rings is 0x00000000.
* The return rings live entirely within the host, so the
* nicaddr field in the RCB isn't used.
*/
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
vrcb = BGE_MEMWIN_START + BGE_RX_RETURN_RING_RCB;
BGE_HOSTADDR(taddr, sc->bge_ldata.bge_rx_return_ring_paddr);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_hi, taddr.bge_addr_hi);
RCB_WRITE_4(sc, vrcb, bge_hostaddr.bge_addr_lo, taddr.bge_addr_lo);
RCB_WRITE_4(sc, vrcb, bge_nicaddr, 0x00000000);
RCB_WRITE_4(sc, vrcb, bge_maxlen_flags,
BGE_RCB_MAXLEN_FLAGS(sc->bge_return_ring_cnt, 0));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Set random backoff seed for TX */
CSR_WRITE_4(sc, BGE_TX_RANDOM_BACKOFF,
- Store pointer to the link-level address right in "struct ifnet" rather than in ifindex_table[]; all (except one) accesses are through ifp anyway. IF_LLADDR() works faster, and all (except one) ifaddr_byindex() users were converted to use ifp->if_addr. - Stop storing a (pointer to) Ethernet address in "struct arpcom", and drop the IFP2ENADDR() macro; all users have been converted to use IF_LLADDR() instead.
2005-11-11 16:04:59 +00:00
IF_LLADDR(sc->bge_ifp)[0] + IF_LLADDR(sc->bge_ifp)[1] +
IF_LLADDR(sc->bge_ifp)[2] + IF_LLADDR(sc->bge_ifp)[3] +
IF_LLADDR(sc->bge_ifp)[4] + IF_LLADDR(sc->bge_ifp)[5] +
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_TX_BACKOFF_SEED_MASK);
/* Set inter-packet gap */
CSR_WRITE_4(sc, BGE_TX_LENGTHS, 0x2620);
/*
* Specify which ring to use for packets that don't match
* any RX rules.
*/
CSR_WRITE_4(sc, BGE_RX_RULES_CFG, 0x08);
/*
* Configure number of RX lists. One interrupt distribution
* list, sixteen active lists, one bad frames class.
*/
CSR_WRITE_4(sc, BGE_RXLP_CFG, 0x181);
/* Inialize RX list placement stats mask. */
CSR_WRITE_4(sc, BGE_RXLP_STATS_ENABLE_MASK, 0x007FFFFF);
CSR_WRITE_4(sc, BGE_RXLP_STATS_CTL, 0x1);
/* Disable host coalescing until we get it set up */
CSR_WRITE_4(sc, BGE_HCC_MODE, 0x00000000);
/* Poll to make sure it's shut down. */
for (i = 0; i < BGE_TIMEOUT; i++) {
if (!(CSR_READ_4(sc, BGE_HCC_MODE) & BGE_HCCMODE_ENABLE))
break;
DELAY(10);
}
if (i == BGE_TIMEOUT) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"host coalescing engine failed to idle\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (ENXIO);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/* 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);
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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 */
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
CSR_WRITE_4(sc, BGE_HCC_STATS_ADDR_HI,
BGE_ADDR_HI(sc->bge_ldata.bge_stats_paddr));
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATS_ADDR_LO,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
BGE_ADDR_LO(sc->bge_ldata.bge_stats_paddr));
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATS_BASEADDR, BGE_STATS_BLOCK);
CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_BASEADDR, BGE_STATUS_BLOCK);
CSR_WRITE_4(sc, BGE_HCC_STATS_TICKS, sc->bge_stat_ticks);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Set up address of status block */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_ADDR_HI,
BGE_ADDR_HI(sc->bge_ldata.bge_status_block_paddr));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_HCC_STATUSBLK_ADDR_LO,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
BGE_ADDR_LO(sc->bge_ldata.bge_status_block_paddr));
sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx = 0;
sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx = 0;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Turn on host coalescing state machine */
CSR_WRITE_4(sc, BGE_HCC_MODE, BGE_HCCMODE_ENABLE);
/* Turn on RX BD completion state machine and enable attentions */
CSR_WRITE_4(sc, BGE_RBDC_MODE,
BGE_RBDCMODE_ENABLE|BGE_RBDCMODE_ATTN);
/* Turn on RX list placement state machine */
CSR_WRITE_4(sc, BGE_RXLP_MODE, BGE_RXLPMODE_ENABLE);
/* Turn on RX list selector state machine. */
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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|
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
((sc->bge_flags & BGE_FLAG_TBI) ?
BGE_PORTMODE_TBI : BGE_PORTMODE_MII));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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 */
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
val = BGE_WDMAMODE_ENABLE|BGE_WDMAMODE_ALL_ATTNS;
/* Enable host coalescing bug fix. */
if (sc->bge_asicrev == BGE_ASICREV_BCM5755 ||
sc->bge_asicrev == BGE_ASICREV_BCM5787)
val |= (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
/* Turn on write DMA state machine */
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
CSR_WRITE_4(sc, BGE_WDMA_MODE, val);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Turn on read DMA state machine */
CSR_WRITE_4(sc, BGE_RDMA_MODE,
BGE_RDMAMODE_ENABLE|BGE_RDMAMODE_ALL_ATTNS);
/* Turn on RX data completion state machine */
CSR_WRITE_4(sc, BGE_RDC_MODE, BGE_RDCMODE_ENABLE);
/* Turn on RX BD initiator state machine */
CSR_WRITE_4(sc, BGE_RBDI_MODE, BGE_RBDIMODE_ENABLE);
/* Turn on RX data and RX BD initiator state machine */
CSR_WRITE_4(sc, BGE_RDBDI_MODE, BGE_RDBDIMODE_ENABLE);
/* Turn on Mbuf cluster free state machine */
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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) */
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_TBI) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_STS, BGE_MISTS_LINK);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
} else {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_SETBIT(sc, BGE_MI_MODE, BGE_MIMODE_AUTOPOLL|10<<16);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
sc->bge_chipid != BGE_CHIPID_BCM5700_B2)
Try to sort out the correct way to generate async link state change interrupts. This is a bit harder than it needs to be because there's more than one way to generate link attentions, at least one of which does not work on the BCM5700, but does on the 5701. For the 5701, we can safely use the 'link changed' bit in the status block, and we enable link change attentions in the mac event register. For the 5700, we have to use MII interrupts, which require checking the MAC status register rather than the status block. This requires doing an extra register access on each interrupt which I'd prefer to avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T shows that we do in fact detect the link going up and down properly on cable insertions/disconnections. Also, avoid twiddling the autopoll enable bit in the MI mode register when doing a PHY read. I think this coupled with the other changes will stop the interrupt storms Paul Saab has been harassing me about. Manually setting the link to 100baseTX full duplex seems to work ok for me. (I'm typing over the 3c996B-T right now.) Lastly, teach the driver how to recognize a 3c996B-SX by checking the hardware config word in the EEPROM in order to detect the media. We attach 5701 fiber cards correctly now, but I haven't verified that they send/receive packets yet since I don't have a second fiber interface at home. (I know that fiber 5700 cards work, so I'm keeping my fingers crossed.)
2002-04-04 06:01:31 +00:00
CSR_WRITE_4(sc, BGE_MAC_EVT_ENB,
BGE_EVTENB_MI_INTERRUPT);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
/*
* Clear any pending link state attention.
* Otherwise some link state change events may be lost until attention
* is cleared by bge_intr() -> bge_link_upd() sequence.
* It's not necessary on newer BCM chips - perhaps enabling link
* state change attentions implies clearing pending attention.
*/
CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED|
BGE_MACSTAT_CFG_CHANGED|BGE_MACSTAT_MI_COMPLETE|
BGE_MACSTAT_LINK_CHANGED);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Enable link state change attentions. */
BGE_SETBIT(sc, BGE_MAC_EVT_ENB, BGE_EVTENB_LINK_CHANGED);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
const struct bge_revision *
bge_lookup_rev(uint32_t chipid)
{
const struct bge_revision *br;
for (br = bge_revisions; br->br_name != NULL; br++) {
if (br->br_chipid == chipid)
return (br);
}
for (br = bge_majorrevs; br->br_name != NULL; br++) {
if (br->br_chipid == BGE_ASICREV(chipid))
return (br);
}
return (NULL);
}
const struct bge_vendor *
bge_lookup_vendor(uint16_t vid)
{
const struct bge_vendor *v;
for (v = bge_vendors; v->v_name != NULL; v++)
if (v->v_id == vid)
return (v);
panic("%s: unknown vendor %d", __func__, vid);
return (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
/*
* Probe for a Broadcom chip. Check the PCI vendor and device IDs
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
* against our list and return its name if we find a match.
*
* Note that since the Broadcom controller contains VPD support, we
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* can get the device name string from the controller itself instead
* of the compiled-in string. This is a little slow, but it guarantees
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
* we'll always announce the right product name. Unfortunately, this
* is possible only later in bge_attach(), when we have established
* access to EEPROM.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_probe(device_t dev)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
struct bge_type *t = bge_devs;
struct bge_softc *sc = device_get_softc(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
bzero(sc, sizeof(struct bge_softc));
sc->bge_dev = dev;
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
while(t->bge_vid != 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 ((pci_get_vendor(dev) == t->bge_vid) &&
(pci_get_device(dev) == t->bge_did)) {
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
char buf[64];
const struct bge_revision *br;
const struct bge_vendor *v;
uint32_t id;
id = pci_read_config(dev, BGE_PCI_MISC_CTL, 4) &
BGE_PCIMISCCTL_ASICREV;
br = bge_lookup_rev(id);
id >>= 16;
v = bge_lookup_vendor(t->bge_vid);
if (br == NULL)
snprintf(buf, 64, "%s unknown ASIC (%#04x)",
v->v_name, id);
else
snprintf(buf, 64, "%s %s, ASIC rev. %#04x",
v->v_name, br->br_name, id);
device_set_desc_copy(dev, buf);
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)
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
sc->bge_flags |= BGE_FLAG_NO3LED;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
t++;
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (ENXIO);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_dma_free(struct bge_softc *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
{
int i;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy DMA maps for RX buffers. */
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
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]);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy DMA maps for jumbo RX buffers. */
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
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]);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy DMA maps for TX buffers. */
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
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);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy standard RX ring. */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_std_ring_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_std_ring_map && sc->bge_ldata.bge_rx_std_ring)
bus_dmamem_free(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_ldata.bge_rx_std_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_rx_std_ring_map);
if (sc->bge_cdata.bge_rx_std_ring_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_rx_std_ring_tag);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy jumbo RX ring. */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_jumbo_ring_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_jumbo_ring_map &&
sc->bge_ldata.bge_rx_jumbo_ring)
bus_dmamem_free(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_ldata.bge_rx_jumbo_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_rx_jumbo_ring_map);
if (sc->bge_cdata.bge_rx_jumbo_ring_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_rx_jumbo_ring_tag);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy RX return ring. */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_return_ring_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_rx_return_ring_tag,
sc->bge_cdata.bge_rx_return_ring_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_rx_return_ring_map &&
sc->bge_ldata.bge_rx_return_ring)
bus_dmamem_free(sc->bge_cdata.bge_rx_return_ring_tag,
sc->bge_ldata.bge_rx_return_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_rx_return_ring_map);
if (sc->bge_cdata.bge_rx_return_ring_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_rx_return_ring_tag);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy TX ring. */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_tx_ring_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_cdata.bge_tx_ring_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_tx_ring_map && sc->bge_ldata.bge_tx_ring)
bus_dmamem_free(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_ldata.bge_tx_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_tx_ring_map);
if (sc->bge_cdata.bge_tx_ring_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_tx_ring_tag);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy status block. */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_status_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_status_map && sc->bge_ldata.bge_status_block)
bus_dmamem_free(sc->bge_cdata.bge_status_tag,
sc->bge_ldata.bge_status_block,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_status_map);
if (sc->bge_cdata.bge_status_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_status_tag);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy statistics block. */
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_stats_map)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_stats_tag,
sc->bge_cdata.bge_stats_map);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (sc->bge_cdata.bge_stats_map && sc->bge_ldata.bge_stats)
bus_dmamem_free(sc->bge_cdata.bge_stats_tag,
sc->bge_ldata.bge_stats,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_cdata.bge_stats_map);
if (sc->bge_cdata.bge_stats_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_stats_tag);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Destroy the parent tag. */
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_cdata.bge_parent_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_parent_tag);
}
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_dma_alloc(device_t dev)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
{
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
struct bge_dmamap_arg ctx;
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_softc *sc;
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
int i, error;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc = device_get_softc(dev);
/*
* Allocate the parent bus DMA tag appropriate for PCI.
*/
add a newbus method for obtaining the bus's bus_dma_tag_t... This is required by arches like sparc64 (not yet implemented) and sun4v where there are seperate IOMMU's for each PCI bus... For all other arches, it will end up returning NULL, which makes it a no-op... Convert a few drivers (the ones we've been working w/ on sun4v) to the new convection... Eventually all drivers will need to replace the parent tag of NULL, w/ bus_get_dma_tag(dev), though dev is usually different for each driver, and will require hand inspection... Reviewed by: scottl (earlier version)
2006-09-03 00:27:42 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(sc->bge_dev),/* parent */
There is no reason to have a PAGE_SIZE alignment enforced on all DMA by the parent tag.
2006-10-19 08:03:22 +00:00
1, 0, /* alignment, boundary */
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_SPACE_MAXADDR, /* lowaddr */
Fix the order of the lowaddr,highaddr arguments in the parent tag. This coincidentally didn't cause any problems, but was definitely wrong.
2005-04-24 02:45:27 +00:00
BUS_SPACE_MAXADDR, /* highaddr */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
NULL, NULL, /* filter, filterarg */
MAXBSIZE, BGE_NSEG_NEW, /* maxsize, nsegments */
BUS_SPACE_MAXSIZE_32BIT,/* maxsegsize */
Don't force busdma to pre-allocate bounce pages for the parent tag.
2004-11-21 04:02:36 +00:00
0, /* flags */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
NULL, NULL, /* lockfunc, lockarg */
&sc->bge_cdata.bge_parent_tag);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (error != 0) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"could not allocate parent dma tag\n");
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
return (ENOMEM);
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/*
* Create tag for RX mbufs.
*/
Use and alignment of 1 instead of ETHER_ALIGN for rx and tx buffers and jumbo frames. BGE hardware with the rx alignment bug will still be handled by the calls to m_adj() that already exist. m_adj() is probably better suited for this task anyways. Just as with if_em, this saves a malloc + several locks per packet and prevents unneeded data copying within busdma.
2004-10-19 02:42:49 +00:00
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag, 1,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
NULL, MCLBYTES * BGE_NSEG_NEW, BGE_NSEG_NEW, MCLBYTES,
BUS_DMA_ALLOCNOW, NULL, NULL, &sc->bge_cdata.bge_mtag);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create DMA maps for RX buffers. */
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
for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
error = bus_dmamap_create(sc->bge_cdata.bge_mtag, 0,
&sc->bge_cdata.bge_rx_std_dmamap[i]);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"can't create DMA map for RX\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
}
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create DMA maps for TX buffers. */
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
for (i = 0; i < BGE_TX_RING_CNT; i++) {
error = bus_dmamap_create(sc->bge_cdata.bge_mtag, 0,
&sc->bge_cdata.bge_tx_dmamap[i]);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"can't create DMA map for RX\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
}
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create tag for standard RX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_STD_RX_RING_SZ, 1, BGE_STD_RX_RING_SZ, 0,
NULL, NULL, &sc->bge_cdata.bge_rx_std_ring_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Allocate DMA'able memory for standard RX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
error = bus_dmamem_alloc(sc->bge_cdata.bge_rx_std_ring_tag,
(void **)&sc->bge_ldata.bge_rx_std_ring, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_rx_std_ring_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_rx_std_ring, BGE_STD_RX_RING_SZ);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Load the address of the standard RX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create tags for jumbo mbufs. */
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
if (BGE_IS_JUMBO_CAPABLE(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
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
Use and alignment of 1 instead of ETHER_ALIGN for rx and tx buffers and jumbo frames. BGE hardware with the rx alignment bug will still be handled by the calls to m_adj() that already exist. m_adj() is probably better suited for this task anyways. Just as with if_em, this saves a malloc + several locks per packet and prevents unneeded data copying within busdma.
2004-10-19 02:42:49 +00:00
1, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
NULL, MJUM9BYTES, BGE_NSEG_JUMBO, PAGE_SIZE,
0, NULL, NULL, &sc->bge_cdata.bge_mtag_jumbo);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
"could not allocate jumbo dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create tag for jumbo RX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_JUMBO_RX_RING_SZ, 1, BGE_JUMBO_RX_RING_SZ, 0,
NULL, NULL, &sc->bge_cdata.bge_rx_jumbo_ring_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
"could not allocate jumbo ring dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Allocate DMA'able memory for jumbo RX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
error = bus_dmamem_alloc(sc->bge_cdata.bge_rx_jumbo_ring_tag,
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
(void **)&sc->bge_ldata.bge_rx_jumbo_ring,
BUS_DMA_NOWAIT | BUS_DMA_ZERO,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
&sc->bge_cdata.bge_rx_jumbo_ring_map);
if (error)
return (ENOMEM);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Load the address of the jumbo RX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create DMA maps for jumbo RX buffers. */
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
for (i = 0; i < BGE_JUMBO_RX_RING_CNT; i++) {
error = bus_dmamap_create(sc->bge_cdata.bge_mtag_jumbo,
0, &sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
"can't create DMA map for jumbo RX\n");
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
}
}
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create tag for RX return ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_RX_RTN_RING_SZ(sc), 1, BGE_RX_RTN_RING_SZ(sc), 0,
NULL, NULL, &sc->bge_cdata.bge_rx_return_ring_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Allocate DMA'able memory for RX return ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
error = bus_dmamem_alloc(sc->bge_cdata.bge_rx_return_ring_tag,
(void **)&sc->bge_ldata.bge_rx_return_ring, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_rx_return_ring_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_rx_return_ring,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
BGE_RX_RTN_RING_SZ(sc));
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Load the address of the RX return ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create tag for TX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_TX_RING_SZ, 1, BGE_TX_RING_SZ, 0, NULL, NULL,
&sc->bge_cdata.bge_tx_ring_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Allocate DMA'able memory for TX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
error = bus_dmamem_alloc(sc->bge_cdata.bge_tx_ring_tag,
(void **)&sc->bge_ldata.bge_tx_ring, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_tx_ring_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_tx_ring, BGE_TX_RING_SZ);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Load the address of the TX ring. */
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create tag for 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
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
NULL, BGE_STATUS_BLK_SZ, 1, BGE_STATUS_BLK_SZ, 0,
NULL, NULL, &sc->bge_cdata.bge_status_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Allocate DMA'able memory for 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
error = bus_dmamem_alloc(sc->bge_cdata.bge_status_tag,
(void **)&sc->bge_ldata.bge_status_block, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_status_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_status_block, BGE_STATUS_BLK_SZ);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Load the address of the 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
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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Create tag for statistics 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
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
PAGE_SIZE, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL,
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
NULL, BGE_STATS_SZ, 1, BGE_STATS_SZ, 0, NULL, NULL,
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
&sc->bge_cdata.bge_stats_tag);
if (error) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "could not allocate dma tag\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
return (ENOMEM);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Allocate DMA'able memory for statistics 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
error = bus_dmamem_alloc(sc->bge_cdata.bge_stats_tag,
(void **)&sc->bge_ldata.bge_stats, BUS_DMA_NOWAIT,
&sc->bge_cdata.bge_stats_map);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (error)
return (ENOMEM);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bzero((char *)sc->bge_ldata.bge_stats, BGE_STATS_SZ);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Load the address of the statstics 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
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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_attach(device_t dev)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
struct ifnet *ifp;
struct bge_softc *sc;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t hwcfg = 0;
uint32_t mac_tmp = 0;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
u_char eaddr[6];
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
int error = 0, rid;
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
int trys, reg;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc = device_get_softc(dev);
sc->bge_dev = dev;
/*
* Map control/status registers.
*/
pci_enable_busmaster(dev);
rid = BGE_PCI_BAR0;
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->bge_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
RF_ACTIVE|PCI_RF_DENSE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (sc->bge_res == NULL) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf (sc->bge_dev, "couldn't map memory\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = ENXIO;
goto fail;
}
sc->bge_btag = rman_get_bustag(sc->bge_res);
sc->bge_bhandle = rman_get_bushandle(sc->bge_res);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Allocate 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
rid = 0;
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->bge_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
RF_SHAREABLE | RF_ACTIVE);
if (sc->bge_irq == NULL) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "couldn't map interrupt\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = ENXIO;
goto fail;
}
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_INIT(sc, device_get_nameunit(dev));
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
/* Save ASIC rev. */
sc->bge_chipid =
pci_read_config(dev, BGE_PCI_MISC_CTL, 4) &
BGE_PCIMISCCTL_ASICREV;
sc->bge_asicrev = BGE_ASICREV(sc->bge_chipid);
sc->bge_chiprev = BGE_CHIPREV(sc->bge_chipid);
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
/* Save chipset family. */
switch (sc->bge_asicrev) {
case BGE_ASICREV_BCM5700:
case BGE_ASICREV_BCM5701:
case BGE_ASICREV_BCM5703:
case BGE_ASICREV_BCM5704:
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
sc->bge_flags |= BGE_FLAG_5700_FAMILY | BGE_FLAG_JUMBO;
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
break;
case BGE_ASICREV_BCM5714_A0:
case BGE_ASICREV_BCM5780:
case BGE_ASICREV_BCM5714:
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
sc->bge_flags |= BGE_FLAG_5714_FAMILY /* | BGE_FLAG_JUMBO */;
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
/* Fall through */
case BGE_ASICREV_BCM5750:
case BGE_ASICREV_BCM5752:
case BGE_ASICREV_BCM5755:
case BGE_ASICREV_BCM5787:
sc->bge_flags |= BGE_FLAG_575X_PLUS;
/* Fall through */
case BGE_ASICREV_BCM5705:
sc->bge_flags |= BGE_FLAG_5705_PLUS;
break;
}
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
/*
* Check if this is a PCI-X or PCI Express device.
*/
#if __FreeBSD_version > 700010
if (pci_find_extcap(dev, PCIY_EXPRESS, &reg) == 0) {
/*
* Found a PCI Express capabilities register, this
* must be a PCI Express device.
*/
if (reg != 0)
sc->bge_flags |= BGE_FLAG_PCIE;
} else if (pci_find_extcap(dev, PCIY_PCIX, &reg) == 0) {
if (reg != 0)
sc->bge_flags |= BGE_FLAG_PCIX;
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
}
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
#else
Use the BGE_IS_* macros consistently. Also add a couple of missing cases to the debug_info function.
2006-12-13 21:03:55 +00:00
if (BGE_IS_5705_PLUS(sc)) {
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
reg = pci_read_config(dev, BGE_PCIE_CAPID_REG, 4);
if ((reg & 0xff) == BGE_PCIE_CAPID)
sc->bge_flags |= BGE_FLAG_PCIE;
} else {
/*
* Check if the device is in PCI-X Mode.
* (This bit is not valid on PCI Express controllers.)
*/
if ((pci_read_config(sc->bge_dev, BGE_PCI_PCISTATE, 4) &
BGE_PCISTATE_PCI_BUSMODE) == 0)
sc->bge_flags |= BGE_FLAG_PCIX;
}
#endif
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +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
/* Try to reset the chip. */
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
if (bge_reset(sc)) {
device_printf(sc->bge_dev, "chip reset failed\n");
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
sc->bge_asf_mode = 0;
Allow the ASF feature to be disabled via a tunable. On one of my systems, bringing up the bge interface results in a complete system freeze when this feature is enabled. Leave it enabled by default.
2006-09-23 18:55:49 +00:00
if (bge_allow_asf && (bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM_SIG)
== BGE_MAGIC_NUMBER)) {
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
if (bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM_NICCFG)
& BGE_HWCFG_ASF) {
sc->bge_asf_mode |= ASF_ENABLE;
sc->bge_asf_mode |= ASF_STACKUP;
if (sc->bge_asicrev == BGE_ASICREV_BCM5750) {
sc->bge_asf_mode |= ASF_NEW_HANDSHAKE;
}
}
}
/* Try to reset the chip again the nice way. */
bge_stop_fw(sc);
bge_sig_pre_reset(sc, BGE_RESET_STOP);
if (bge_reset(sc)) {
device_printf(sc->bge_dev, "chip reset failed\n");
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
bge_sig_legacy(sc, BGE_RESET_STOP);
bge_sig_post_reset(sc, BGE_RESET_STOP);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 (bge_chipinit(sc)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "chip initialization failed\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
/*
* Get station address from the EEPROM.
*/
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
mac_tmp = bge_readmem_ind(sc, 0x0c14);
if ((mac_tmp >> 16) == 0x484b) {
eaddr[0] = (u_char)(mac_tmp >> 8);
eaddr[1] = (u_char)mac_tmp;
mac_tmp = bge_readmem_ind(sc, 0x0c18);
eaddr[2] = (u_char)(mac_tmp >> 24);
eaddr[3] = (u_char)(mac_tmp >> 16);
eaddr[4] = (u_char)(mac_tmp >> 8);
eaddr[5] = (u_char)mac_tmp;
} else if (bge_read_eeprom(sc, eaddr,
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_EE_MAC_OFFSET + 2, ETHER_ADDR_LEN)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "failed to read station address\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/* 5705 limits RX return ring to 512 entries. */
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (BGE_IS_5705_PLUS(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
sc->bge_return_ring_cnt = BGE_RETURN_RING_CNT_5705;
else
sc->bge_return_ring_cnt = BGE_RETURN_RING_CNT;
if (bge_dma_alloc(dev)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev,
"failed to allocate DMA resources\n");
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Set default tuneable values. */
sc->bge_stat_ticks = BGE_TICKS_PER_SEC;
sc->bge_rx_coal_ticks = 150;
sc->bge_tx_coal_ticks = 150;
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
sc->bge_rx_max_coal_bds = 10;
sc->bge_tx_max_coal_bds = 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
/* Set up ifnet structure */
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp = if_alloc(IFT_ETHER);
if (ifp == NULL) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "failed to if_alloc()\n");
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_softc = sc;
Replace the if_name and if_unit members of struct ifnet with new members if_xname, if_dname, and if_dunit. if_xname is the name of the interface and if_dname/unit are the driver name and instance. This change paves the way for interface renaming and enhanced pseudo device creation and configuration symantics. Approved By: re (in principle) Reviewed By: njl, imp Tested On: i386, amd64, sparc64 Obtained From: NetBSD (if_xname)
2003-10-31 18:32:15 +00:00
if_initname(ifp, device_get_name(dev), device_get_unit(dev));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
ifp->if_ioctl = bge_ioctl;
ifp->if_start = bge_start;
ifp->if_init = bge_init;
ifp->if_mtu = ETHERMTU;
Add altq(4) support.
2004-10-30 21:21:10 +00:00
ifp->if_snd.ifq_drv_maxlen = BGE_TX_RING_CNT - 1;
IFQ_SET_MAXLEN(&ifp->if_snd, ifp->if_snd.ifq_drv_maxlen);
IFQ_SET_READY(&ifp->if_snd);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_hwassist = BGE_CSUM_FEATURES;
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
ifp->if_capabilities = IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING |
This driver can do hardware VLAN tagging + checksum offloading. In collaboration with: Mihail Balikov <mihail.balikov interbgc.com>
2006-01-30 13:45:55 +00:00
IFCAP_VLAN_MTU | IFCAP_VLAN_HWCSUM;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_capenable = ifp->if_capabilities;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
ifp->if_capabilities |= IFCAP_POLLING;
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Whitespace.
2006-06-08 10:19:16 +00:00
/*
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
* 5700 B0 chips do not support checksumming correctly due
* to hardware bugs.
*/
if (sc->bge_chipid == BGE_CHIPID_BCM5700_B0) {
ifp->if_capabilities &= ~IFCAP_HWCSUM;
ifp->if_capenable &= IFCAP_HWCSUM;
ifp->if_hwassist = 0;
}
Try to sort out the correct way to generate async link state change interrupts. This is a bit harder than it needs to be because there's more than one way to generate link attentions, at least one of which does not work on the BCM5700, but does on the 5701. For the 5701, we can safely use the 'link changed' bit in the status block, and we enable link change attentions in the mac event register. For the 5700, we have to use MII interrupts, which require checking the MAC status register rather than the status block. This requires doing an extra register access on each interrupt which I'd prefer to avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T shows that we do in fact detect the link going up and down properly on cable insertions/disconnections. Also, avoid twiddling the autopoll enable bit in the MI mode register when doing a PHY read. I think this coupled with the other changes will stop the interrupt storms Paul Saab has been harassing me about. Manually setting the link to 100baseTX full duplex seems to work ok for me. (I'm typing over the 3c996B-T right now.) Lastly, teach the driver how to recognize a 3c996B-SX by checking the hardware config word in the EEPROM in order to detect the media. We attach 5701 fiber cards correctly now, but I haven't verified that they send/receive packets yet since I don't have a second fiber interface at home. (I know that fiber 5700 cards work, so I'm keeping my fingers crossed.)
2002-04-04 06:01:31 +00:00
/*
* Figure out what sort of media we have by checking the
Obtain the media type from the shared memory and only use the eeprom as a fallback.
2003-02-05 08:54:36 +00:00
* hardware config word in the first 32k of NIC internal memory,
* or fall back to examining the EEPROM if necessary.
* Note: on some BCM5700 cards, this value appears to be unset.
* If that's the case, we have to rely on identifying the NIC
* by its PCI subsystem ID, as we do below for the SysKonnect
* SK-9D41.
Try to sort out the correct way to generate async link state change interrupts. This is a bit harder than it needs to be because there's more than one way to generate link attentions, at least one of which does not work on the BCM5700, but does on the 5701. For the 5701, we can safely use the 'link changed' bit in the status block, and we enable link change attentions in the mac event register. For the 5700, we have to use MII interrupts, which require checking the MAC status register rather than the status block. This requires doing an extra register access on each interrupt which I'd prefer to avoid, but them's the breaks. Testing with both a 3c996-T and 3c996B-T shows that we do in fact detect the link going up and down properly on cable insertions/disconnections. Also, avoid twiddling the autopoll enable bit in the MI mode register when doing a PHY read. I think this coupled with the other changes will stop the interrupt storms Paul Saab has been harassing me about. Manually setting the link to 100baseTX full duplex seems to work ok for me. (I'm typing over the 3c996B-T right now.) Lastly, teach the driver how to recognize a 3c996B-SX by checking the hardware config word in the EEPROM in order to detect the media. We attach 5701 fiber cards correctly now, but I haven't verified that they send/receive packets yet since I don't have a second fiber interface at home. (I know that fiber 5700 cards work, so I'm keeping my fingers crossed.)
2002-04-04 06:01:31 +00:00
*/
Obtain the media type from the shared memory and only use the eeprom as a fallback.
2003-02-05 08:54:36 +00:00
if (bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM_SIG) == BGE_MAGIC_NUMBER)
hwcfg = bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM_NICCFG);
else {
Fix bge_eeprom_getbyte() to return 1 when timeout happens. Previously it always returned 0 which means success regardless of EEPROM status. While here, add a check whether EEPROM read is successful. Submitted by: jkim
2005-12-22 02:03:57 +00:00
if (bge_read_eeprom(sc, (caddr_t)&hwcfg, BGE_EE_HWCFG_OFFSET,
sizeof(hwcfg))) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "failed to read EEPROM\n");
Fix bge_eeprom_getbyte() to return 1 when timeout happens. Previously it always returned 0 which means success regardless of EEPROM status. While here, add a check whether EEPROM read is successful. Submitted by: jkim
2005-12-22 02:03:57 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Obtain the media type from the shared memory and only use the eeprom as a fallback.
2003-02-05 08:54:36 +00:00
hwcfg = ntohl(hwcfg);
}
if ((hwcfg & BGE_HWCFG_MEDIA) == BGE_MEDIA_FIBER)
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
sc->bge_flags |= BGE_FLAG_TBI;
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
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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)
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
sc->bge_flags |= BGE_FLAG_TBI;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_TBI) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
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 {
/*
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
* Do transceiver setup and tell the firmware the
* driver is down so we can try to get access the
* probe if ASF is running. Retry a couple of times
* if we get a conflict with the ASF firmware accessing
* the PHY.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
BGE_CLRBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
again:
bge_asf_driver_up(sc);
trys = 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 (mii_phy_probe(dev, &sc->bge_miibus,
bge_ifmedia_upd, bge_ifmedia_sts)) {
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
if (trys++ < 4) {
device_printf(sc->bge_dev, "Try again\n");
bge_miibus_writereg(sc->bge_dev, 1, MII_BMCR, BMCR_RESET);
goto again;
}
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "MII without any PHY!\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_release_resources(sc);
error = ENXIO;
goto fail;
}
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
/*
* Now tell the firmware we are going up after probing the PHY
*/
if (sc->bge_asf_mode & ASF_STACKUP)
BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
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.
*/
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5701 &&
sc->bge_flags & BGE_FLAG_PCIX)
sc->bge_flags |= BGE_FLAG_RX_ALIGNBUG;
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Call MI attach routine.
*/
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ether_ifattach(ifp, eaddr);
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
callout_init_mtx(&sc->bge_stat_ch, &sc->bge_mtx, 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
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
/*
* Hookup IRQ last.
*/
error = bus_setup_intr(dev, sc->bge_irq, INTR_TYPE_NET | INTR_MPSAFE,
bge_intr, sc, &sc->bge_intrhand);
if (error) {
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
bge_detach(dev);
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "couldn't set up irq\n");
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
bge_add_sysctls(sc);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
fail:
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (error);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_detach(device_t dev)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
struct bge_softc *sc;
struct ifnet *ifp;
sc = device_get_softc(dev);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING)
ether_poll_deregister(ifp);
#endif
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_stop(sc);
bge_reset(sc);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
ether_ifdetach(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_TBI) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifmedia_removeall(&sc->bge_ifmedia);
} else {
bus_generic_detach(dev);
device_delete_child(dev, sc->bge_miibus);
}
bge_release_resources(sc);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_release_resources(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
{
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
device_t dev;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
dev = sc->bge_dev;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (sc->bge_vpd_prodname != NULL)
free(sc->bge_vpd_prodname, M_DEVBUF);
if (sc->bge_vpd_readonly != NULL)
free(sc->bge_vpd_readonly, M_DEVBUF);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (sc->bge_intrhand != NULL)
bus_teardown_intr(dev, sc->bge_irq, sc->bge_intrhand);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (sc->bge_irq != NULL)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bus_release_resource(dev, SYS_RES_IRQ, 0, sc->bge_irq);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
if (sc->bge_res != NULL)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bus_release_resource(dev, SYS_RES_MEMORY,
BGE_PCI_BAR0, sc->bge_res);
Fix "struct ifnet" leaks when attach() fails in the middle.
2005-09-16 11:25:19 +00:00
if (sc->bge_ifp != NULL)
if_free(sc->bge_ifp);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bge_dma_free(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
if (mtx_initialized(&sc->bge_mtx)) /* XXX */
BGE_LOCK_DESTROY(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_reset(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
{
device_t dev;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t cachesize, command, pcistate, reset;
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
void (*write_op)(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
int i, val = 0;
dev = sc->bge_dev;
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
if (BGE_IS_5705_PLUS(sc) && !BGE_IS_5714_FAMILY(sc))
if (sc->bge_flags & BGE_FLAG_PCIE)
write_op = bge_writemem_direct;
else
write_op = bge_writemem_ind;
else
write_op = bge_writereg_ind;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Save some important PCI state. */
cachesize = pci_read_config(dev, BGE_PCI_CACHESZ, 4);
command = pci_read_config(dev, BGE_PCI_CMD, 4);
pcistate = pci_read_config(dev, BGE_PCI_PCISTATE, 4);
pci_write_config(dev, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_INDIRECT_ACCESS|BGE_PCIMISCCTL_MASK_PCI_INTR|
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
BGE_HIF_SWAP_OPTIONS|BGE_PCIMISCCTL_PCISTATE_RW, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
/* Disable fastboot on controllers that support it. */
if (sc->bge_asicrev == BGE_ASICREV_BCM5752 ||
sc->bge_asicrev == BGE_ASICREV_BCM5755 ||
sc->bge_asicrev == BGE_ASICREV_BCM5787) {
if (bootverbose)
device_printf(sc->bge_dev, "%s: Disabling fastboot\n",
__FUNCTION__);
CSR_WRITE_4(sc, BGE_FASTBOOT_PC, 0x0);
}
/*
* Write the magic number to SRAM at offset 0xB50.
* When firmware finishes its initialization it will
* write ~BGE_MAGIC_NUMBER to the same location.
*/
bge_writemem_ind(sc, BGE_SOFTWARE_GENCOMM, BGE_MAGIC_NUMBER);
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. */
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_PCIE) {
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 (CSR_READ_4(sc, 0x7e2c) == 0x60) /* PCIE 1.0 */
CSR_WRITE_4(sc, 0x7e2c, 0x20);
if (sc->bge_chipid != BGE_CHIPID_BCM5750_A0) {
/* Prevent PCIE link training during global reset */
CSR_WRITE_4(sc, BGE_MISC_CFG, (1<<29));
reset |= (1<<29);
}
}
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
/*
* Set GPHY Power Down Override to leave GPHY
* powered up in D0 uninitialized.
*/
Use the BGE_IS_* macros consistently. Also add a couple of missing cases to the debug_info function.
2006-12-13 21:03:55 +00:00
if (BGE_IS_5705_PLUS(sc))
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
reset |= 0x04000000;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 */
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
write_op(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. */
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_PCIE) {
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_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);
}
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
/* Set PCIE max payload size to 128 bytes and clear error status. */
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
pci_write_config(dev, 0xd8, 0xf5000, 4);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Reset some of the PCI state that got zapped by 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
pci_write_config(dev, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_INDIRECT_ACCESS|BGE_PCIMISCCTL_MASK_PCI_INTR|
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
BGE_HIF_SWAP_OPTIONS|BGE_PCIMISCCTL_PCISTATE_RW, 4);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
pci_write_config(dev, BGE_PCI_CACHESZ, cachesize, 4);
pci_write_config(dev, BGE_PCI_CMD, command, 4);
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
write_op(sc, BGE_MISC_CFG, (65 << 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
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. */
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
if (BGE_IS_5714_FAMILY(sc)) {
uint32_t val;
val = CSR_READ_4(sc, BGE_MARB_MODE);
CSR_WRITE_4(sc, BGE_MARB_MODE, BGE_MARBMODE_ENABLE | val);
} else
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
/*
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
* Poll until we see the 1's complement of the magic number.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* This indicates that the firmware initialization
* is complete.
*/
for (i = 0; i < BGE_TIMEOUT; i++) {
val = bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM);
if (val == ~BGE_MAGIC_NUMBER)
break;
DELAY(10);
}
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (i == BGE_TIMEOUT) {
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
device_printf(sc->bge_dev, "firmware handshake timed out! "
"found 0x%08X\n", val);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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 support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
if (sc->bge_flags & BGE_FLAG_PCIE) {
reset = bge_readmem_ind(sc, 0x7c00);
bge_writemem_ind(sc, 0x7c00, reset | (1 << 25));
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Fix up byte swapping. */
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
CSR_WRITE_4(sc, BGE_MODE_CTL, BGE_DMA_SWAP_OPTIONS|
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_MODECTL_BYTESWAP_DATA);
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
/* Tell the ASF firmware we are up */
if (sc->bge_asf_mode & ASF_STACKUP)
BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_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.
*/
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5704 &&
sc->bge_flags & BGE_FLAG_TBI) {
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
uint32_t serdescfg;
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
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
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. */
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_PCIE &&
sc->bge_chipid != BGE_CHIPID_BCM5750_A0) {
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
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);
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
return(0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* Frame reception handling. This is called if there's a frame
* on the receive return list.
*
* Note: we have to be able to handle two possibilities here:
A big rewrite of receive Jumbo frame handling. Remove the local Jumbo cluster allocator, that wasn't MPSAFE. Instead, utilize our new generic UMA jumbo cluster allocator. Since UMA gives us a 9k piece that is contigous in virtual memory, but isn't contigous in physical memory we need to handle a few segments. To deal with this we utilize Tigon chip feature - extended RX descriptors, that can handle up to four DMA segments for one frame. Details: o Remove bge_alloc_jumbo_mem(), bge_free_jumbo_mem(), bge_jalloc(), bge_jfree() functions. o Remove SLIST heads, bge_jumbo_tag, bge_jumbo_map from softc. o Use extended RX BDs for Jumbo receive producer ring, and initialize it appropriately. o New bge_newbuf_jumbo(): - Allocate an mbuf with Jumbo cluster with help of m_cljget(). - Load the cluster for DMA with help of bus_dmamap_load_mbuf_sg(). - Assert that we got 3 segments in the DMA mapping. - Fill in these 3 segments into the extended RX descriptor.
2005-12-08 16:11:45 +00:00
* 1) the frame is from the jumbo receive ring
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* 2) the frame is from the standard receive ring
*/
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_rxeof(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;
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);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Nothing to do. */
Optimize bge_rxeof() & bge_txeof(): return immediately if there are no packets to process. It could give us [significant?] perfomance increase if there is big difference between RX/TX flows. Submitted by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 15:16:03 +00:00
if (sc->bge_rx_saved_considx ==
sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx)
return;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_return_ring_tag,
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
sc->bge_cdata.bge_rx_return_ring_map, BUS_DMASYNC_POSTREAD);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_POSTREAD);
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
if (BGE_IS_JUMBO_CAPABLE(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_rx_jumbo_ring_tag,
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
sc->bge_cdata.bge_rx_jumbo_ring_map, BUS_DMASYNC_POSTREAD);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t 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
struct mbuf *m = NULL;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint16_t 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
int have_tag = 0;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING) {
if (sc->rxcycles <= 0)
break;
sc->rxcycles--;
}
#endif
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
cur_rx =
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
&sc->bge_ldata.bge_rx_return_ring[sc->bge_rx_saved_considx];
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
rxidx = cur_rx->bge_idx;
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
BGE_INC(sc->bge_rx_saved_considx, sc->bge_return_ring_cnt);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Do not strip VLAN tag in promiscuous mode.
2006-09-18 22:18:21 +00:00
if (!(ifp->if_flags & IFF_PROMISC) &&
(cur_rx->bge_flags & BGE_RXBDFLAG_VLAN_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
have_tag = 1;
vlan_tag = cur_rx->bge_vlan_tag;
}
if (cur_rx->bge_flags & BGE_RXBDFLAG_JUMBO_RING) {
BGE_INC(sc->bge_jumbo, BGE_JUMBO_RX_RING_CNT);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[rxidx],
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[rxidx]);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m = sc->bge_cdata.bge_rx_jumbo_chain[rxidx];
sc->bge_cdata.bge_rx_jumbo_chain[rxidx] = NULL;
jumbocnt++;
if (cur_rx->bge_flags & BGE_RXBDFLAG_ERROR) {
ifp->if_ierrors++;
bge_newbuf_jumbo(sc, sc->bge_jumbo, m);
continue;
}
if (bge_newbuf_jumbo(sc,
sc->bge_jumbo, NULL) == ENOBUFS) {
ifp->if_ierrors++;
bge_newbuf_jumbo(sc, sc->bge_jumbo, m);
continue;
}
} else {
BGE_INC(sc->bge_std, BGE_STD_RX_RING_CNT);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[rxidx],
BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_rx_std_dmamap[rxidx]);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m = sc->bge_cdata.bge_rx_std_chain[rxidx];
sc->bge_cdata.bge_rx_std_chain[rxidx] = NULL;
stdcnt++;
if (cur_rx->bge_flags & BGE_RXBDFLAG_ERROR) {
ifp->if_ierrors++;
bge_newbuf_std(sc, sc->bge_std, m);
continue;
}
if (bge_newbuf_std(sc, sc->bge_std,
NULL) == ENOBUFS) {
ifp->if_ierrors++;
bge_newbuf_std(sc, sc->bge_std, m);
continue;
}
}
ifp->if_ipackets++;
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
#ifndef __NO_STRICT_ALIGNMENT
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
/*
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
* For architectures with strict alignment we must make sure
* the payload is aligned.
Work around what appears to be a chip bug in the BCM5701 that shows up when operating in PCI-X mode. For some received packets there is data corruption in the first few bytes in that case. Aligning the packet buffer eliminates the corruption. With this fix, the code that offsets the packet buffer up by 2 bytes to align the payload is disabled for BCM5701s operating in PCI-X mode. On the i386, which permits unaligned accesses, the payload is left unaligned. On other platforms, the packet is copied after reception to force alignment of the payload. Obviously, this work-around reduces performance in those cases (BCM5701 plus PCI-X) where it is in effect. MFC after: 3 days
2002-06-24 22:04:15 +00:00
*/
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_RX_ALIGNBUG) {
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
bcopy(m->m_data, m->m_data + ETHER_ALIGN,
cur_rx->bge_len);
m->m_data += ETHER_ALIGN;
}
#endif
By not setting No_CRC in the Mode Control Register, we must also reduce the size of the packet by 4 bytes to remove the ethernet crc. Submitted by: jdp
2003-09-23 19:54:32 +00:00
m->m_pkthdr.len = m->m_len = cur_rx->bge_len - ETHER_CRC_LEN;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m->m_pkthdr.rcvif = ifp;
Bring the advertised interface capabilities into line with the reality (in particular, bge(4) hasn't supported rxcsum since if_bge.c#1.5) Clean up some aspects of capabilities usage, i.e. stop using if_hwassist to see whether we are doing offload now because if_hwassist is for TCP/IP layer and it is subordinate to if_capenable. Thanks to: Aled Morris for donating a nice bge(4) NIC to me Reviewed by: -net, -hackers (silence)
2004-07-24 13:45:38 +00:00
if (ifp->if_capenable & IFCAP_RXCSUM) {
Check for BGE_RXBDFLAG_IP_CSUM flag before marking mbuf as with valid checksum. This fixes operation of PPPoE and other non-IP Ethernet protocols.
2006-01-21 09:54:32 +00:00
if (cur_rx->bge_flags & BGE_RXBDFLAG_IP_CSUM) {
m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED;
if ((cur_rx->bge_ip_csum ^ 0xffff) == 0)
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
}
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
if (cur_rx->bge_flags & BGE_RXBDFLAG_TCP_UDP_CSUM &&
m->m_pkthdr.len >= ETHER_MIN_NOPAD) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
m->m_pkthdr.csum_data =
cur_rx->bge_tcp_udp_csum;
Enable 'complete' rx checksum offloading (i.e. let chip calculate checksums with pseudo header for tcp/udp packets). This could save one in_pseudo() call per incoming tcp/udp packet. Approved by: glebius (mentor) MFC after: 3 weeks
2006-02-02 09:58:31 +00:00
m->m_pkthdr.csum_flags |=
CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
}
/*
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
* If we received a packet with a vlan tag,
* attach that information to the packet.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
- Fix VLAN_INPUT_TAG() macro, so that it doesn't touch mtag in case if memory allocation failed. - Remove fourth argument from VLAN_INPUT_TAG(), that was used incorrectly in almost all drivers. Indicate failure with mbuf value of NULL. In collaboration with: yongari, ru, sam
2005-12-18 18:24:27 +00:00
if (have_tag) {
Move ethernet VLAN tags from mtags to its own mbuf packet header field m_pkthdr.ether_vlan. The presence of the M_VLANTAG flag on the mbuf signifies the presence and validity of its content. Drivers that support hardware VLAN tag stripping fill in the received VLAN tag (containing both vlan and priority information) into the ether_vtag mbuf packet header field: m->m_pkthdr.ether_vtag = vlan_id; /* ntohs()? */ m->m_flags |= M_VLANTAG; to mark the packet m with the specified VLAN tag. On output the driver should check the mbuf for the M_VLANTAG flag to see if a VLAN tag is present and valid: if (m->m_flags & M_VLANTAG) { ... = m->m_pkthdr.ether_vtag; /* htons()? */ ... pass tag to hardware ... } VLAN tags are stored in host byte order. Byte swapping may be necessary. (Note: This driver conversion was mechanic and did not add or remove any byte swapping in the drivers.) Remove zone_mtag_vlan UMA zone and MTAG_VLAN definition. No more tag memory allocation have to be done. Reviewed by: thompsa, yar Sponsored by: TCP/IP Optimization Fundraise 2005
2006-09-17 13:33:30 +00:00
m->m_pkthdr.ether_vtag = vlan_tag;
m->m_flags |= M_VLANTAG;
- Fix VLAN_INPUT_TAG() macro, so that it doesn't touch mtag in case if memory allocation failed. - Remove fourth argument from VLAN_INPUT_TAG(), that was used incorrectly in almost all drivers. Indicate failure with mbuf value of NULL. In collaboration with: yongari, ru, sam
2005-12-18 18:24:27 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
(*ifp->if_input)(ifp, m);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
if (stdcnt > 0)
bus_dmamap_sync(sc->bge_cdata.bge_rx_std_ring_tag,
sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_PREWRITE);
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
if (BGE_IS_JUMBO_CAPABLE(sc) && jumbocnt > 0)
bus_dmamap_sync(sc->bge_cdata.bge_rx_jumbo_ring_tag,
sc->bge_cdata.bge_rx_jumbo_ring_map, BUS_DMASYNC_PREWRITE);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MBX_RX_CONS0_LO, sc->bge_rx_saved_considx);
if (stdcnt)
CSR_WRITE_4(sc, BGE_MBX_RX_STD_PROD_LO, sc->bge_std);
if (jumbocnt)
CSR_WRITE_4(sc, BGE_MBX_RX_JUMBO_PROD_LO, sc->bge_jumbo);
- Correct collision counter for BCM5705+. This register is read/clear. - Correct RX packet drop counter for BCM5705+. This register is read/clear and it wraps very quickly under heavy packet drops because only the lower ten bits are valid according to the documentation. However, it seems few more bits are actually valid and the rest bits are always zeros[1]. Therefore, we don't mask them off here. To get accurate packet drop count, we need to check the register from bge_rxeof(). It is commented out for now, not to penalize normal operation. Actual performance impact should be measured later. - Correct integer casting from u_long to uint32_t. Casting is not really needed for all supported platforms but we better do this correctly[2]. Tested by: bde[1] Suggested by: bde[2]
2006-12-11 18:00:34 +00:00
#ifdef notyet
/*
* This register wraps very quickly under heavy packet drops.
* If you need correct statistics, you can enable this check.
*/
if (BGE_IS_5705_PLUS(sc))
ifp->if_ierrors += CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_DROPS);
#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 void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_txeof(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 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);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Nothing to do. */
Optimize bge_rxeof() & bge_txeof(): return immediately if there are no packets to process. It could give us [significant?] perfomance increase if there is big difference between RX/TX flows. Submitted by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 15:16:03 +00:00
if (sc->bge_tx_saved_considx ==
sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx)
return;
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_cdata.bge_tx_ring_map,
BUS_DMASYNC_POSTREAD|BUS_DMASYNC_POSTWRITE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Go through our tx ring and free mbufs for those
* frames that have been sent.
*/
while (sc->bge_tx_saved_considx !=
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx) {
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t 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
idx = sc->bge_tx_saved_considx;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
cur_tx = &sc->bge_ldata.bge_tx_ring[idx];
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (cur_tx->bge_flags & BGE_TXBDFLAG_END)
ifp->if_opackets++;
if (sc->bge_cdata.bge_tx_chain[idx] != NULL) {
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[idx],
BUS_DMASYNC_POSTWRITE);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_mtag,
sc->bge_cdata.bge_tx_dmamap[idx]);
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
m_freem(sc->bge_cdata.bge_tx_chain[idx]);
sc->bge_cdata.bge_tx_chain[idx] = NULL;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
sc->bge_txcnt--;
BGE_INC(sc->bge_tx_saved_considx, BGE_TX_RING_CNT);
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
sc->bge_timer = 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 (cur_tx != NULL)
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
static void
bge_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
{
struct bge_softc *sc = ifp->if_softc;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t statusword;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
BGE_LOCK(sc);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
BGE_UNLOCK(sc);
return;
}
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_POSTREAD);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
statusword = atomic_readandclear_32(
&sc->bge_ldata.bge_status_block->bge_status);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_PREREAD);
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
/* Note link event. It will be processed by POLL_AND_CHECK_STATUS cmd */
if (statusword & BGE_STATFLAG_LINKSTATE_CHANGED)
sc->bge_link_evt++;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
if (cmd == POLL_AND_CHECK_STATUS)
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if ((sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
sc->bge_chipid != BGE_CHIPID_BCM5700_B2) ||
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
sc->bge_link_evt || (sc->bge_flags & BGE_FLAG_TBI))
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
bge_link_upd(sc);
Since bge_rxeof() & bge_txeof() depends on status block data it should be synchronized on every call of bge_poll_locked(). Suggested by: Mihail Balikov <mihail.balikov AT interbgc DOT com> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:41:08 +00:00
sc->rxcycles = count;
bge_rxeof(sc);
bge_txeof(sc);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
bge_start_locked(ifp);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
BGE_UNLOCK(sc);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
}
#endif /* DEVICE_POLLING */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_intr(void *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
{
struct bge_softc *sc;
struct ifnet *ifp;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
uint32_t statusword;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc = xsc;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
ifp = sc->bge_ifp;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
if (ifp->if_capenable & IFCAP_POLLING) {
BGE_UNLOCK(sc);
return;
}
#endif
Fix the interrupt handler to do the mandatory PCI flush before looking at DMA memory. The could contribute towards missed link state changes under heavy bus load.
2006-04-15 08:13:06 +00:00
/*
* Do the mandatory PCI flush as well as get the link status.
*/
statusword = CSR_READ_4(sc, BGE_MAC_STS) & 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
Fix the interrupt handler to do the mandatory PCI flush before looking at DMA memory. The could contribute towards missed link state changes under heavy bus load.
2006-04-15 08:13:06 +00:00
/* Ack interrupt and stop others from occuring. */
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 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
Fix the interrupt handler to do the mandatory PCI flush before looking at DMA memory. The could contribute towards missed link state changes under heavy bus load.
2006-04-15 08:13:06 +00:00
/* Make sure the descriptor ring indexes are coherent. */
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_POSTREAD);
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map, BUS_DMASYNC_PREREAD);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if ((sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
sc->bge_chipid != BGE_CHIPID_BCM5700_B2) ||
Fix the interrupt handler to do the mandatory PCI flush before looking at DMA memory. The could contribute towards missed link state changes under heavy bus load.
2006-04-15 08:13:06 +00:00
statusword || sc->bge_link_evt)
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
bge_link_upd(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Check RX return ring producer/consumer. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_rxeof(sc);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Check TX ring producer/consumer. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_txeof(sc);
}
/* Re-enable interrupts. */
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 0);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
bge_start_locked(ifp);
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
static void
bge_asf_driver_up(struct bge_softc *sc)
{
if (sc->bge_asf_mode & ASF_STACKUP) {
/* Send ASF heartbeat aprox. every 2s */
if (sc->bge_asf_count)
sc->bge_asf_count --;
else {
sc->bge_asf_count = 5;
bge_writemem_ind(sc, BGE_SOFTWARE_GENCOMM_FW,
BGE_FW_DRV_ALIVE);
bge_writemem_ind(sc, BGE_SOFTWARE_GENNCOMM_FW_LEN, 4);
bge_writemem_ind(sc, BGE_SOFTWARE_GENNCOMM_FW_DATA, 3);
CSR_WRITE_4(sc, BGE_CPU_EVENT,
CSR_READ_4(sc, BGE_CPU_EVENT) != (1 << 14));
}
}
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
bge_tick(void *xsc)
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
{
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
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
struct mii_data *mii = NULL;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (BGE_IS_5705_PLUS(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
bge_stats_update_regs(sc);
else
bge_stats_update(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if ((sc->bge_flags & BGE_FLAG_TBI) == 0) {
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
mii = device_get_softc(sc->bge_miibus);
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
/* Don't mess with the PHY in IPMI/ASF mode */
if (!((sc->bge_asf_mode & ASF_STACKUP) && (sc->bge_link)))
mii_tick(mii);
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
} else {
/*
* Since in TBI mode auto-polling can't be used we should poll
* link status manually. Here we register pending link event
* and trigger interrupt.
*/
#ifdef DEVICE_POLLING
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* In polling mode we poll link state in bge_poll(). */
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
if (!(sc->bge_ifp->if_capenable & IFCAP_POLLING))
#endif
{
sc->bge_link_evt++;
BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_INTR_SET);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
bge_asf_driver_up(sc);
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
bge_watchdog(sc);
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
callout_reset(&sc->bge_stat_ch, hz, bge_tick, sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_stats_update_regs(struct bge_softc *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
{
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
struct ifnet *ifp;
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
- Correct collision counter for BCM5705+. This register is read/clear. - Correct RX packet drop counter for BCM5705+. This register is read/clear and it wraps very quickly under heavy packet drops because only the lower ten bits are valid according to the documentation. However, it seems few more bits are actually valid and the rest bits are always zeros[1]. Therefore, we don't mask them off here. To get accurate packet drop count, we need to check the register from bge_rxeof(). It is commented out for now, not to penalize normal operation. Actual performance impact should be measured later. - Correct integer casting from u_long to uint32_t. Casting is not really needed for all supported platforms but we better do this correctly[2]. Tested by: bde[1] Suggested by: bde[2]
2006-12-11 18:00:34 +00:00
ifp->if_collisions += CSR_READ_4(sc, BGE_MAC_STATS +
Simplify statistics updates, remove redundant register reads, and add discarded RX packets to input error for BCM5705 or newer chipset as the others. Unfortunately we cannot do the same for output errors because ifOutDiscards equivalent register does not exist. While I am here, replace misleading and wrong BGE_RX_STATS/BGE_TX_STATS with BGE_MAC_STATS. They were reversed but worked accidently.
2006-12-01 01:08:52 +00:00
offsetof(struct bge_mac_stats_regs, etherStatsCollisions));
- Correct collision counter for BCM5705+. This register is read/clear. - Correct RX packet drop counter for BCM5705+. This register is read/clear and it wraps very quickly under heavy packet drops because only the lower ten bits are valid according to the documentation. However, it seems few more bits are actually valid and the rest bits are always zeros[1]. Therefore, we don't mask them off here. To get accurate packet drop count, we need to check the register from bge_rxeof(). It is commented out for now, not to penalize normal operation. Actual performance impact should be measured later. - Correct integer casting from u_long to uint32_t. Casting is not really needed for all supported platforms but we better do this correctly[2]. Tested by: bde[1] Suggested by: bde[2]
2006-12-11 18:00:34 +00:00
ifp->if_ierrors += CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_DROPS);
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_stats_update(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;
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
bus_size_t stats;
Simplify statistics updates, remove redundant register reads, and add discarded RX packets to input error for BCM5705 or newer chipset as the others. Unfortunately we cannot do the same for output errors because ifOutDiscards equivalent register does not exist. While I am here, replace misleading and wrong BGE_RX_STATS/BGE_TX_STATS with BGE_MAC_STATS. They were reversed but worked accidently.
2006-12-01 01:08:52 +00:00
uint32_t cnt; /* current register value */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
stats = BGE_MEMWIN_START + BGE_STATS_BLOCK;
#define READ_STAT(sc, stats, stat) \
CSR_READ_4(sc, stats + offsetof(struct bge_stats, stat))
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Use one counter instead of four to make up TX collision stats as BCM5705+. It reduces chance of errors from multiple counter wraps at the same time.
2006-12-13 17:28:51 +00:00
cnt = READ_STAT(sc, stats, txstats.etherStatsCollisions.bge_addr_lo);
- Correct collision counter for BCM5705+. This register is read/clear. - Correct RX packet drop counter for BCM5705+. This register is read/clear and it wraps very quickly under heavy packet drops because only the lower ten bits are valid according to the documentation. However, it seems few more bits are actually valid and the rest bits are always zeros[1]. Therefore, we don't mask them off here. To get accurate packet drop count, we need to check the register from bge_rxeof(). It is commented out for now, not to penalize normal operation. Actual performance impact should be measured later. - Correct integer casting from u_long to uint32_t. Casting is not really needed for all supported platforms but we better do this correctly[2]. Tested by: bde[1] Suggested by: bde[2]
2006-12-11 18:00:34 +00:00
ifp->if_collisions += (uint32_t)(cnt - sc->bge_tx_collisions);
- Count packets discarded by RX/TX MAC (cause of FIFO overflow, etc) as input/output interface errors. - Keep values of rx/tx discards & tx collisions inside struct bge_softc. So we can keep statistic across ifconfig down/up runs (cause bringing bge up will reset chip). Approved by: glebius (mentor) MFC after: 1 week
2006-01-17 23:01:58 +00:00
sc->bge_tx_collisions = cnt;
cnt = READ_STAT(sc, stats, ifInDiscards.bge_addr_lo);
- Correct collision counter for BCM5705+. This register is read/clear. - Correct RX packet drop counter for BCM5705+. This register is read/clear and it wraps very quickly under heavy packet drops because only the lower ten bits are valid according to the documentation. However, it seems few more bits are actually valid and the rest bits are always zeros[1]. Therefore, we don't mask them off here. To get accurate packet drop count, we need to check the register from bge_rxeof(). It is commented out for now, not to penalize normal operation. Actual performance impact should be measured later. - Correct integer casting from u_long to uint32_t. Casting is not really needed for all supported platforms but we better do this correctly[2]. Tested by: bde[1] Suggested by: bde[2]
2006-12-11 18:00:34 +00:00
ifp->if_ierrors += (uint32_t)(cnt - sc->bge_rx_discards);
- Count packets discarded by RX/TX MAC (cause of FIFO overflow, etc) as input/output interface errors. - Keep values of rx/tx discards & tx collisions inside struct bge_softc. So we can keep statistic across ifconfig down/up runs (cause bringing bge up will reset chip). Approved by: glebius (mentor) MFC after: 1 week
2006-01-17 23:01:58 +00:00
sc->bge_rx_discards = cnt;
cnt = READ_STAT(sc, stats, txstats.ifOutDiscards.bge_addr_lo);
- Correct collision counter for BCM5705+. This register is read/clear. - Correct RX packet drop counter for BCM5705+. This register is read/clear and it wraps very quickly under heavy packet drops because only the lower ten bits are valid according to the documentation. However, it seems few more bits are actually valid and the rest bits are always zeros[1]. Therefore, we don't mask them off here. To get accurate packet drop count, we need to check the register from bge_rxeof(). It is commented out for now, not to penalize normal operation. Actual performance impact should be measured later. - Correct integer casting from u_long to uint32_t. Casting is not really needed for all supported platforms but we better do this correctly[2]. Tested by: bde[1] Suggested by: bde[2]
2006-12-11 18:00:34 +00:00
ifp->if_oerrors += (uint32_t)(cnt - sc->bge_tx_discards);
- Count packets discarded by RX/TX MAC (cause of FIFO overflow, etc) as input/output interface errors. - Keep values of rx/tx discards & tx collisions inside struct bge_softc. So we can keep statistic across ifconfig down/up runs (cause bringing bge up will reset chip). Approved by: glebius (mentor) MFC after: 1 week
2006-01-17 23:01:58 +00:00
sc->bge_tx_discards = cnt;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 1/2). - Give up endianess support and switch to native-endian format for accessing hardware structures. In fact embedded processor for BCM57xx is big-endian architure(MIPS) and it requires native-endian format for NIC structures.The NIC performs necessary byte/word swapping depending on programmed endian type. - With above changes all htole16/htole32 calls were gone. - Remove bge_vhandle member in softc and changed to use explicit register access. This may add additional performance penalty that than that of previous memory access. But most of the access is performed on initialization phase(e.g. RCB setup), it would be negligible. Due to incorrect use of bus_dma(9) in bge(4) it still panics sparc64 system in device detach path. The issue would be fixed in next patch. Reviewed by: jkim (initial version) Silence from: ps Tested by: glebius Obtained from: NetBSD via OpenBSD
2005-12-15 05:48:49 +00:00
#undef READ_STAT
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
/*
* Pad outbound frame to ETHER_MIN_NOPAD for an unusual reason.
* The bge hardware will pad out Tx runts to ETHER_MIN_NOPAD,
* but when such padded frames employ the bge IP/TCP checksum offload,
* the hardware checksum assist gives incorrect results (possibly
* from incorporating its own padding into the UDP/TCP checksum; who knows).
* If we pad such runts with zeros, the onboard checksum comes out correct.
*/
static __inline int
bge_cksum_pad(struct mbuf *m)
{
int padlen = ETHER_MIN_NOPAD - m->m_pkthdr.len;
struct mbuf *last;
/* If there's only the packet-header and we can pad there, use it. */
if (m->m_pkthdr.len == m->m_len && M_WRITABLE(m) &&
M_TRAILINGSPACE(m) >= padlen) {
last = m;
} else {
/*
* Walk packet chain to find last mbuf. We will either
* pad there, or append a new mbuf and pad it.
*/
for (last = m; last->m_next != NULL; last = last->m_next);
if (!(M_WRITABLE(last) && M_TRAILINGSPACE(last) >= padlen)) {
/* Allocate new empty mbuf, pad it. Compact later. */
struct mbuf *n;
MGET(n, M_DONTWAIT, MT_DATA);
if (n == NULL)
return (ENOBUFS);
n->m_len = 0;
last->m_next = n;
last = n;
}
}
/* Now zero the pad area, to avoid the bge cksum-assist bug. */
memset(mtod(last, caddr_t) + last->m_len, 0, padlen);
last->m_len += padlen;
m->m_pkthdr.len += padlen;
return (0);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Encapsulate an mbuf chain in the tx ring by coupling the mbuf data
* pointers to descriptors.
*/
static int
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
bge_encap(struct bge_softc *sc, struct mbuf **m_head, uint32_t *txidx)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
bus_dma_segment_t segs[BGE_NSEG_NEW];
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
bus_dmamap_t map;
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
struct bge_tx_bd *d;
struct mbuf *m = *m_head;
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
uint32_t idx = *txidx;
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
uint16_t csum_flags;
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
int nsegs, i, error;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_cksum_pad() can modify the mbuf, thus call it before bus_dmamap_load(). Submitted by: yongari
2006-08-18 13:53:53 +00:00
csum_flags = 0;
if (m->m_pkthdr.csum_flags) {
if (m->m_pkthdr.csum_flags & CSUM_IP)
csum_flags |= BGE_TXBDFLAG_IP_CSUM;
if (m->m_pkthdr.csum_flags & (CSUM_TCP | CSUM_UDP)) {
csum_flags |= BGE_TXBDFLAG_TCP_UDP_CSUM;
if (m->m_pkthdr.len < ETHER_MIN_NOPAD &&
(error = bge_cksum_pad(m)) != 0) {
m_freem(m);
*m_head = NULL;
return (error);
}
}
if (m->m_flags & M_LASTFRAG)
csum_flags |= BGE_TXBDFLAG_IP_FRAG_END;
else if (m->m_flags & M_FRAG)
csum_flags |= BGE_TXBDFLAG_IP_FRAG;
}
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
map = sc->bge_cdata.bge_tx_dmamap[idx];
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_mtag, map, m, segs,
&nsegs, BUS_DMA_NOWAIT);
if (error == EFBIG) {
m = m_defrag(m, M_DONTWAIT);
if (m == NULL) {
m_freem(*m_head);
*m_head = NULL;
return (ENOBUFS);
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
}
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
*m_head = m;
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_mtag, map, m,
segs, &nsegs, BUS_DMA_NOWAIT);
if (error) {
m_freem(m);
*m_head = NULL;
Whitespace.
2006-06-08 10:19:16 +00:00
return (error);
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
}
} else if (error != 0)
return (error);
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
* Sanity check: avoid coming within 16 descriptors
* of the end of the ring.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
if (nsegs > (BGE_TX_RING_CNT - sc->bge_txcnt - 16)) {
bus_dmamap_unload(sc->bge_cdata.bge_mtag, map);
return (ENOBUFS);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Add bge(4) support for big-endian architectures(part 2/2). - removed unused funtion bge_handle_events(). - removed bus_dmamap_destroy(9) calls for DMA maps created by bus_dmamem_alloc(9). This should fix panics seen on sparc64 in device detach. - added check for parent DMA tag creation. - switched to use __NO_STRICT_ALIGNMENT as bge(4) supports all architectures. - added missing bus_dmamap_sync(9) in bge_txeof(). - added missing bus_dmamap_sync(9) in bge_encap(). - corrected memory synchronization operation on status block. As the driver just read status block that was DMAed by NIC it should use BUS_DMASYNC_POSTREAD. Likewise the driver does not need to write status block back, so remove unnecessary bus_dmamap_sync(9) calls in bge_intr(). - corrected memory synchronization operation on RX return ring. The driver only read the block so remove unnecessary bus_dmamap_sync(9) in bge_rxeof(). - force bus_dmamap_sync(9) for only modified descriptors. Blindly synching all desciptor rings would reduce performance. - call bus_dmamap_sync(9) for DMA maps that were modified in bge_rxeof(). Reviewed by: jkim(initial version) Tested by: glebius(i386), jkim(amd64 initial version)
2005-12-22 01:44:27 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag, map, BUS_DMASYNC_PREWRITE);
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
for (i = 0; ; i++) {
d = &sc->bge_ldata.bge_tx_ring[idx];
d->bge_addr.bge_addr_lo = BGE_ADDR_LO(segs[i].ds_addr);
d->bge_addr.bge_addr_hi = BGE_ADDR_HI(segs[i].ds_addr);
d->bge_len = segs[i].ds_len;
d->bge_flags = csum_flags;
if (i == nsegs - 1)
break;
BGE_INC(idx, BGE_TX_RING_CNT);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
/* Mark the last segment as end of packet... */
d->bge_flags |= BGE_TXBDFLAG_END;
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
/* ... and put VLAN tag into first segment. */
d = &sc->bge_ldata.bge_tx_ring[*txidx];
Move ethernet VLAN tags from mtags to its own mbuf packet header field m_pkthdr.ether_vlan. The presence of the M_VLANTAG flag on the mbuf signifies the presence and validity of its content. Drivers that support hardware VLAN tag stripping fill in the received VLAN tag (containing both vlan and priority information) into the ether_vtag mbuf packet header field: m->m_pkthdr.ether_vtag = vlan_id; /* ntohs()? */ m->m_flags |= M_VLANTAG; to mark the packet m with the specified VLAN tag. On output the driver should check the mbuf for the M_VLANTAG flag to see if a VLAN tag is present and valid: if (m->m_flags & M_VLANTAG) { ... = m->m_pkthdr.ether_vtag; /* htons()? */ ... pass tag to hardware ... } VLAN tags are stored in host byte order. Byte swapping may be necessary. (Note: This driver conversion was mechanic and did not add or remove any byte swapping in the drivers.) Remove zone_mtag_vlan UMA zone and MTAG_VLAN definition. No more tag memory allocation have to be done. Reviewed by: thompsa, yar Sponsored by: TCP/IP Optimization Fundraise 2005
2006-09-17 13:33:30 +00:00
if (m->m_flags & M_VLANTAG) {
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
d->bge_flags |= BGE_TXBDFLAG_VLAN_TAG;
Move ethernet VLAN tags from mtags to its own mbuf packet header field m_pkthdr.ether_vlan. The presence of the M_VLANTAG flag on the mbuf signifies the presence and validity of its content. Drivers that support hardware VLAN tag stripping fill in the received VLAN tag (containing both vlan and priority information) into the ether_vtag mbuf packet header field: m->m_pkthdr.ether_vtag = vlan_id; /* ntohs()? */ m->m_flags |= M_VLANTAG; to mark the packet m with the specified VLAN tag. On output the driver should check the mbuf for the M_VLANTAG flag to see if a VLAN tag is present and valid: if (m->m_flags & M_VLANTAG) { ... = m->m_pkthdr.ether_vtag; /* htons()? */ ... pass tag to hardware ... } VLAN tags are stored in host byte order. Byte swapping may be necessary. (Note: This driver conversion was mechanic and did not add or remove any byte swapping in the drivers.) Remove zone_mtag_vlan UMA zone and MTAG_VLAN definition. No more tag memory allocation have to be done. Reviewed by: thompsa, yar Sponsored by: TCP/IP Optimization Fundraise 2005
2006-09-17 13:33:30 +00:00
d->bge_vlan_tag = m->m_pkthdr.ether_vtag;
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
} else
d->bge_vlan_tag = 0;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/*
* Insure that the map for this transmission
* is placed at the array index of the last descriptor
* in this chain.
*/
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
sc->bge_cdata.bge_tx_dmamap[*txidx] = sc->bge_cdata.bge_tx_dmamap[idx];
sc->bge_cdata.bge_tx_dmamap[idx] = map;
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
sc->bge_cdata.bge_tx_chain[idx] = m;
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
sc->bge_txcnt += nsegs;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
BGE_INC(idx, BGE_TX_RING_CNT);
*txidx = idx;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* Main transmit routine. To avoid having to do mbuf copies, we put pointers
* to the mbuf data regions directly in the transmit descriptors.
*/
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_start_locked(struct ifnet *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 bge_softc *sc;
struct mbuf *m_head = NULL;
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
uint32_t prodidx;
With ALTQ, it is possible that although the queue was not empty when we entered the interface start function, no packets were actually dequeued. Therefore, keep a count of how many packets we really added onto the tx chain, and initiate a transmit only if the count is non-zero.
2004-10-30 22:59:30 +00:00
int count = 0;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
sc = ifp->if_softc;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
if (!sc->bge_link || IFQ_DRV_IS_EMPTY(&ifp->if_snd))
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
prodidx = sc->bge_tx_prodidx;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
while(sc->bge_cdata.bge_tx_chain[prodidx] == NULL) {
Add altq(4) support.
2004-10-30 21:21:10 +00:00
IFQ_DRV_DEQUEUE(&ifp->if_snd, m_head);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (m_head == NULL)
break;
/*
Bring the advertised interface capabilities into line with the reality (in particular, bge(4) hasn't supported rxcsum since if_bge.c#1.5) Clean up some aspects of capabilities usage, i.e. stop using if_hwassist to see whether we are doing offload now because if_hwassist is for TCP/IP layer and it is subordinate to if_capenable. Thanks to: Aled Morris for donating a nice bge(4) NIC to me Reviewed by: -net, -hackers (silence)
2004-07-24 13:45:38 +00:00
* XXX
* The code inside the if() block is never reached since we
* must mark CSUM_IP_FRAGS in our if_hwassist to start getting
* requests to checksum TCP/UDP in a fragmented packet.
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
*
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
* XXX
* safety overkill. If this is a fragmented packet chain
* with delayed TCP/UDP checksums, then only encapsulate
* it if we have enough descriptors to handle the entire
* chain at once.
* (paranoia -- may not actually be needed)
*/
if (m_head->m_flags & M_FIRSTFRAG &&
m_head->m_pkthdr.csum_flags & (CSUM_DELAY_DATA)) {
if ((BGE_TX_RING_CNT - sc->bge_txcnt) <
m_head->m_pkthdr.csum_data + 16) {
Add altq(4) support.
2004-10-30 21:21:10 +00:00
IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
break;
}
}
/*
* Pack the data into the transmit ring. If we
* don't have room, set the OACTIVE flag and wait
* for the NIC to drain the ring.
*/
Rewrite bge_encap() so that it takes mbuf ** argument. In this case if m_defrag(9) changes pointer we can show this to our caller. Reviewed by: yongari
2006-08-17 09:53:04 +00:00
if (bge_encap(sc, &m_head, &prodidx)) {
if (m_head == NULL)
break;
Add altq(4) support.
2004-10-30 21:21:10 +00:00
IFQ_DRV_PREPEND(&ifp->if_snd, m_head);
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
break;
}
With ALTQ, it is possible that although the queue was not empty when we entered the interface start function, no packets were actually dequeued. Therefore, keep a count of how many packets we really added onto the tx chain, and initiate a transmit only if the count is non-zero.
2004-10-30 22:59:30 +00:00
++count;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* If there's a BPF listener, bounce a copy of this frame
* to him.
*/
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
BPF_MTAP(ifp, m_head);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
if (count == 0)
/* No packets were dequeued. */
With ALTQ, it is possible that although the queue was not empty when we entered the interface start function, no packets were actually dequeued. Therefore, keep a count of how many packets we really added onto the tx chain, and initiate a transmit only if the count is non-zero.
2004-10-30 22:59:30 +00:00
return;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Transmit. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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, prodidx);
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
/* 5700 b2 errata */
- Change the full Asic revision defines to CHIPID to better since the ASIC revision is really the major number of the CHIPID. Also store the chipid, asic rev and chip revision in the softc for later use. - The write twice to send producer index workaround only applies to the 5700_BX chips, so only do it there. Requested by: jdp - Do not initalize the LED's to 0x00. The default configuration the chip comes up in should yeild proper operation of the LED's. Confirmed by: John Cagle <john.cagle@hp.com> Approved by: re (blanket)
2003-05-07 21:51:13 +00:00
if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
Implement the write twice to send producer index workaround for BCM5700 chips as described in the Broadcom Errata documents. Obtained from: NetBSD and Broadcom Errata documentation
2003-05-04 00:07:21 +00:00
CSR_WRITE_4(sc, BGE_MBX_TX_HOST_PROD0_LO, prodidx);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Since BGE_MBX_TX_HOST_PROD0_LO register is write-only to software, we can cache its value in the softc. Eliminates one PCI register write per call to bge_start(). A 1.8% speedup for UDP_RR test on my old box. Obtained from: NetBSD(jonathan) via delphij
2005-12-18 20:26:12 +00:00
sc->bge_tx_prodidx = prodidx;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/*
* Set a timeout in case the chip goes out to lunch.
*/
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
sc->bge_timer = 5;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_start(struct ifnet *ifp)
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
{
struct bge_softc *sc;
sc = ifp->if_softc;
BGE_LOCK(sc);
bge_start_locked(ifp);
BGE_UNLOCK(sc);
}
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_init_locked(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;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint16_t *m;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_ASSERT(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
/* Cancel pending I/O and flush buffers. */
bge_stop(sc);
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
bge_stop_fw(sc);
bge_sig_pre_reset(sc, BGE_RESET_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_reset(sc);
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
bge_sig_legacy(sc, BGE_RESET_START);
bge_sig_post_reset(sc, BGE_RESET_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_chipinit(sc);
/*
* Init the various state machines, ring
* control blocks and firmware.
*/
if (bge_blockinit(sc)) {
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf(sc->bge_dev, "initialization failure\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
return;
}
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Specify MTU. */
CSR_WRITE_4(sc, BGE_RX_MTU, ifp->if_mtu +
Set the BGE_RX_MTU register correctly so that we can receive slightly larger than normal frames, to account for the case where a bge(4) NIC is used with VLANs. Since we set the IFCAP_VLAN_MTU flag, we must allow reception of frames up to 1522 bytes in size rather than 1518. Note that it is possible to work around this bug by doing: # ifconfig bge0 mtu 1504 prior to configuring any VLAN interfaces.
2003-08-04 05:50:53 +00:00
ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Load our MAC address. */
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
m = (uint16_t *)IF_LLADDR(sc->bge_ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
CSR_WRITE_4(sc, BGE_MAC_ADDR1_LO, htons(m[0]));
CSR_WRITE_4(sc, BGE_MAC_ADDR1_HI, (htons(m[1]) << 16) | htons(m[2]));
Do not strip VLAN tag in promiscuous mode.
2006-09-18 22:18:21 +00:00
/* Program promiscuous mode. */
bge_setpromisc(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
/* 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) {
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t v, i;
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
for (i = 0; i < 10; i++) {
DELAY(20);
v = bge_readmem_ind(sc, BGE_STD_RX_RINGS + 8);
if (v == (MCLBYTES - ETHER_ALIGN))
break;
}
if (i == 10)
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
device_printf (sc->bge_dev,
"5705 A0 chip failed to load RX ring\n");
Add support for the BCM5705 and its ilk. Changes: - 5705 doesn't support jumbo frames - Statistics must be read from registers - RX return ring must be capped at 512 entries - Omit initialization of certain device blocks - Acknowledge link change interrupts by setting the 'link changed' bit in the status register (used to have no effect) - Remember to toggle the MI completion bit too - Set the mbuf low watermark differently (on-chip memory buffers, not BSD mbufs) - Don't enable Ethernet@WireSpeed feature for certain 5705 chip revs - Add additional PCI IDs for 5705 and 5782 parts - Add a forgotten 5704 PCI ID Most changes ripped kicking and screaming from the Broadcom linux driver. Thanks to Paul Saab for sanity testing. (My lack of sanity has been confirmed.)
2003-07-16 00:09:56 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* Init jumbo RX ring. */
if (ifp->if_mtu > (ETHERMTU + ETHER_HDR_LEN + ETHER_CRC_LEN))
bge_init_rx_ring_jumbo(sc);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Init our RX return ring index. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_rx_saved_considx = 0;
Simplify statistics updates, remove redundant register reads, and add discarded RX packets to input error for BCM5705 or newer chipset as the others. Unfortunately we cannot do the same for output errors because ifOutDiscards equivalent register does not exist. While I am here, replace misleading and wrong BGE_RX_STATS/BGE_TX_STATS with BGE_MAC_STATS. They were reversed but worked accidently.
2006-12-01 01:08:52 +00:00
/* Init our RX/TX stat counters. */
sc->bge_rx_discards = sc->bge_tx_discards = sc->bge_tx_collisions = 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
/* Init TX ring. */
bge_init_tx_ring(sc);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Turn on transmitter. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_TX_MODE, BGE_TXMODE_ENABLE);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Turn on receiver. */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_RX_MODE, BGE_RXMODE_ENABLE);
/* Tell firmware we're alive. */
BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
/* Disable interrupts if we are polling. */
if (ifp->if_capenable & IFCAP_POLLING) {
BGE_SETBIT(sc, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 1);
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 1);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 1);
} else
#endif
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +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
/* Enable host interrupts. */
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
{
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
BGE_SETBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_CLEAR_INTA);
BGE_CLRBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 0);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
}
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +00:00
bge_ifmedia_upd_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
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags |= IFF_DRV_RUNNING;
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
callout_reset(&sc->bge_stat_ch, hz, bge_tick, sc);
}
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_init(void *xsc)
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
{
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
}
/*
* Set media options.
*/
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_ifmedia_upd(struct ifnet *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
{
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +00:00
struct bge_softc *sc = ifp->if_softc;
int res;
BGE_LOCK(sc);
res = bge_ifmedia_upd_locked(ifp);
BGE_UNLOCK(sc);
return (res);
}
static int
bge_ifmedia_upd_locked(struct ifnet *ifp)
{
struct bge_softc *sc = ifp->if_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
struct mii_data *mii;
struct ifmedia *ifm;
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +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
ifm = &sc->bge_ifmedia;
/* If this is a 1000baseX NIC, enable the TBI port. */
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_TBI) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (EINVAL);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
switch(IFM_SUBTYPE(ifm->ifm_media)) {
case IFM_AUTO:
Backout the change I made before 5.4-R since I wasn't aware that it was only a problem with one particular switch module. Create a kernel option BGE_FAKE_AUTONEG that restores the 5.4 behavior, which should make the DNLK switch module work. IBM/Intel blades with Intel or AD switch modules should work without patching or kernel options with this commit. Hardware for testing provided by several folks, including Danny Braniss <danny@cs.huji.ac.il>, Achim Patzner <ap@bnc.net>, and OffMyServer. Approved by: re
2005-06-24 21:43:47 +00:00
/*
* The BCM5704 ASIC appears to have a special
* mechanism for programming the autoneg
* advertisement registers in TBI mode.
*/
make BGE_FAKE_AUTONEG a tunable. This allows one to change the behavior of the driver pre-boot. NOTE: This patch was made for DragonFly BSD by Sepherosa Ziehau. PR: kern/94833 Submitted by: Devon H. O'Dell Obtained from: DragonFly MFC after: 1 month
2006-04-25 15:56:52 +00:00
if (bge_fake_autoneg == 0 &&
sc->bge_asicrev == BGE_ASICREV_BCM5704) {
Backout the change I made before 5.4-R since I wasn't aware that it was only a problem with one particular switch module. Create a kernel option BGE_FAKE_AUTONEG that restores the 5.4 behavior, which should make the DNLK switch module work. IBM/Intel blades with Intel or AD switch modules should work without patching or kernel options with this commit. Hardware for testing provided by several folks, including Danny Braniss <danny@cs.huji.ac.il>, Achim Patzner <ap@bnc.net>, and OffMyServer. Approved by: re
2005-06-24 21:43:47 +00:00
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:
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (EINVAL);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
1) Ignore link events for MII/GMII cards if MI auto-polling disabled. This should fix strange link state behaviour reported for bcm5721 & bcm5704c 2) Clear bge_link flag in bge_stop() 3) Force link state check after bge_ifmedia_upd(). Otherwise we can miss link event if PHY changes it's state fast enough. Tested by: phk (bcm5704c) Approved by: glebius (mentor) MFC after: 1 week
2006-02-17 14:33:35 +00:00
sc->bge_link_evt++;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
mii = device_get_softc(sc->bge_miibus);
if (mii->mii_instance) {
struct mii_softc *miisc;
for (miisc = LIST_FIRST(&mii->mii_phys); miisc != NULL;
miisc = LIST_NEXT(miisc, mii_list))
mii_phy_reset(miisc);
}
mii_mediachg(mii);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (0);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* Report current media status.
*/
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +00:00
struct bge_softc *sc = ifp->if_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
struct mii_data *mii;
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +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
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_TBI) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
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;
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
else {
ifmr->ifm_active |= IFM_NONE;
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +00:00
BGE_UNLOCK(sc);
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
return;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifmr->ifm_active |= IFM_1000_SX;
if (CSR_READ_4(sc, BGE_MAC_MODE) & BGE_MACMODE_HALF_DUPLEX)
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
ifmr->ifm_active |= IFM_HDX;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
else
ifmr->ifm_active |= IFM_FDX;
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +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;
}
mii = device_get_softc(sc->bge_miibus);
mii_pollstat(mii);
ifmr->ifm_active = mii->mii_media_active;
ifmr->ifm_status = mii->mii_media_status;
Properly lock ifmedia callbacks. This should prevent concurrent access to PHY. Following issues should be resolved: - random watchdog timeouts (caused by concurrent phy access) - some link state issues - non working TX if media type was set explicitly PR: kern/98738 Approved by: glebius (mentor) MFC after: 2 weeks
2006-08-24 14:41:16 +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
}
static int
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_ioctl(struct ifnet *ifp, u_long command, caddr_t 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
{
struct bge_softc *sc = ifp->if_softc;
struct ifreq *ifr = (struct ifreq *) data;
struct mii_data *mii;
Simplify promiscuous mode check and make ALLMULTI independent from it.
2006-09-18 20:54:40 +00:00
int flags, 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
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
switch (command) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
case SIOCSIFMTU:
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
if (ifr->ifr_mtu < ETHERMIN ||
((BGE_IS_JUMBO_CAPABLE(sc)) &&
ifr->ifr_mtu > BGE_JUMBO_MTU) ||
((!BGE_IS_JUMBO_CAPABLE(sc)) &&
ifr->ifr_mtu > ETHERMTU))
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
else if (ifp->if_mtu != ifr->ifr_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_mtu = ifr->ifr_mtu;
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_init(sc);
}
break;
case SIOCSIFFLAGS:
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
if (ifp->if_flags & IFF_UP) {
/*
* If only the state of the PROMISC flag changed,
* then just use the 'set promisc mode' command
* instead of reinitializing the entire NIC. Doing
* a full re-init means reloading the firmware and
* waiting for it to start up, which may take a
Setting IFF_ALLMULTI on a running interface didn't call bge_setmulti() as intended, resulting in a non-working multicast routing. Fix it. MFC after: 3 days
2006-02-01 10:11:24 +00:00
* second or two. Similarly for ALLMULTI.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
Simplify promiscuous mode check and make ALLMULTI independent from it.
2006-09-18 20:54:40 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
flags = ifp->if_flags ^ sc->bge_if_flags;
Do not strip VLAN tag in promiscuous mode.
2006-09-18 22:18:21 +00:00
if (flags & IFF_PROMISC)
bge_setpromisc(sc);
Simplify promiscuous mode check and make ALLMULTI independent from it.
2006-09-18 20:54:40 +00:00
if (flags & IFF_ALLMULTI)
bge_setmulti(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
} else
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
bge_init_locked(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
} else {
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_stop(sc);
}
}
sc->bge_if_flags = ifp->if_flags;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = 0;
break;
case SIOCADDMULTI:
case SIOCDELMULTI:
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_setmulti(sc);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
error = 0;
}
break;
case SIOCSIFMEDIA:
case SIOCGIFMEDIA:
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_TBI) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 = ifmedia_ioctl(ifp, ifr,
&sc->bge_ifmedia, command);
} else {
mii = device_get_softc(sc->bge_miibus);
error = ifmedia_ioctl(ifp, ifr,
&mii->mii_media, command);
}
break;
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
case SIOCSIFCAP:
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
mask = ifr->ifr_reqcap ^ ifp->if_capenable;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
#ifdef DEVICE_POLLING
if (mask & IFCAP_POLLING) {
if (ifr->ifr_reqcap & IFCAP_POLLING) {
error = ether_poll_register(bge_poll, ifp);
if (error)
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (error);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
BGE_LOCK(sc);
BGE_SETBIT(sc, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 1);
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 1);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 1);
Whitespace.
2006-06-08 10:19:16 +00:00
ifp->if_capenable |= IFCAP_POLLING;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
BGE_UNLOCK(sc);
} else {
error = ether_poll_deregister(ifp);
/* Enable interrupt even in error case */
BGE_LOCK(sc);
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 0);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 0);
BGE_CLRBIT(sc, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_MASK_PCI_INTR);
CSR_WRITE_4(sc, BGE_MBX_IRQ0_LO, 0);
ifp->if_capenable &= ~IFCAP_POLLING;
BGE_UNLOCK(sc);
}
}
#endif
- All bge(4) supported hardware is known to support RX/TX checksum offloading, except for BGE_CHIPID_BCM5700_B0, which is buggy. - All bge(4) supported hardware, has a bug that produces incorrect checksums on Ethernet runts. However, in case of a transmitted packet, the latter can be padded with zeroes, and the checksum would be correct. (Probably chip includes the pad data into checksum). In case of receive, we just don't trust checksum data in received runts. Obtained from: NetBSD (jonathan) via Mihail Balikov
2005-12-22 15:14:42 +00:00
if (mask & IFCAP_HWCSUM) {
ifp->if_capenable ^= IFCAP_HWCSUM;
if (IFCAP_HWCSUM & ifp->if_capenable &&
IFCAP_HWCSUM & ifp->if_capabilities)
Bring the advertised interface capabilities into line with the reality (in particular, bge(4) hasn't supported rxcsum since if_bge.c#1.5) Clean up some aspects of capabilities usage, i.e. stop using if_hwassist to see whether we are doing offload now because if_hwassist is for TCP/IP layer and it is subordinate to if_capenable. Thanks to: Aled Morris for donating a nice bge(4) NIC to me Reviewed by: -net, -hackers (silence)
2004-07-24 13:45:38 +00:00
ifp->if_hwassist = BGE_CSUM_FEATURES;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
else
Bring the advertised interface capabilities into line with the reality (in particular, bge(4) hasn't supported rxcsum since if_bge.c#1.5) Clean up some aspects of capabilities usage, i.e. stop using if_hwassist to see whether we are doing offload now because if_hwassist is for TCP/IP layer and it is subordinate to if_capenable. Thanks to: Aled Morris for donating a nice bge(4) NIC to me Reviewed by: -net, -hackers (silence)
2004-07-24 13:45:38 +00:00
ifp->if_hwassist = 0;
This driver can do hardware VLAN tagging + checksum offloading. In collaboration with: Mihail Balikov <mihail.balikov interbgc.com>
2006-01-30 13:45:55 +00:00
VLAN_CAPABILITIES(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
break;
default:
network interface driver changes: o don't strip the Ethernet header from inbound packets; pass packets up the stack intact (required significant changes to some drivers) o reference common definitions in net/ethernet.h (e.g. ETHER_ALIGN) o track ether_ifattach/ether_ifdetach API changes o track bpf changes (use BPF_TAP and BPF_MTAP) o track vlan changes (ifnet capabilities, revised processing scheme, etc.) o use if_input to pass packets "up" o call ether_ioctl for default handling of ioctls Reviewed by: many Approved by: re
2002-11-14 23:54:55 +00:00
error = ether_ioctl(ifp, command, data);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
break;
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
return (error);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
static void
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
bge_watchdog(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
{
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
struct ifnet *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
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
BGE_LOCK_ASSERT(sc);
if (sc->bge_timer == 0 || --sc->bge_timer)
return;
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Use device_printf() and if_printf() rather than printf() and axe bge_unit from the softc. Requested by: marius
2005-12-23 02:04:41 +00:00
if_printf(ifp, "watchdog timeout -- resetting\n");
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Propagate rename of IFF_OACTIVE and IFF_RUNNING to IFF_DRV_OACTIVE and IFF_DRV_RUNNING, as well as the move from ifnet.if_flags to ifnet.if_drv_flags. Device drivers are now responsible for synchronizing access to these flags, as they are in if_drv_flags. This helps prevent races between the network stack and device driver in maintaining the interface flags field. Many __FreeBSD__ and __FreeBSD_version checks maintained and continued; some less so. Reviewed by: pjd, bz MFC after: 7 days
2005-08-09 10:20:02 +00:00
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
Fix my error in rev. 1.152 Submitted by: oleg
2006-12-04 14:35:17 +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
ifp->if_oerrors++;
}
/*
* Stop the adapter and free any mbufs allocated to the
* RX and TX lists.
*/
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_stop(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;
struct ifmedia_entry *ifm;
struct mii_data *mii = NULL;
int mtmp, itmp;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK_ASSERT(sc);
Stop embedding struct ifnet at the top of driver softcs. Instead the struct ifnet or the layer 2 common structure it was embedded in have been replaced with a struct ifnet pointer to be filled by a call to the new function, if_alloc(). The layer 2 common structure is also allocated via if_alloc() based on the interface type. It is hung off the new struct ifnet member, if_l2com. This change removes the size of these structures from the kernel ABI and will allow us to better manage them as interfaces come and go. Other changes of note: - Struct arpcom is no longer referenced in normal interface code. Instead the Ethernet address is accessed via the IFP2ENADDR() macro. To enforce this ac_enaddr has been renamed to _ac_enaddr. - The second argument to ether_ifattach is now always the mac address from driver private storage rather than sometimes being ac_enaddr. Reviewed by: sobomax, sam
2005-06-10 16:49:24 +00:00
ifp = sc->bge_ifp;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if ((sc->bge_flags & BGE_FLAG_TBI) == 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
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
/*
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
* Disable all of the receiver blocks.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_RX_MODE, BGE_RXMODE_ENABLE);
BGE_CLRBIT(sc, BGE_RBDI_MODE, BGE_RBDIMODE_ENABLE);
BGE_CLRBIT(sc, BGE_RXLP_MODE, BGE_RXLPMODE_ENABLE);
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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);
/*
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
* Disable all of the transmit blocks.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_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);
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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);
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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);
- Add BGE_FLAG_JUMBO flag which idicates jumbo frame capability. Some day we may be able to support jumbo frames for BCM5714 and BCM5780. - Rename BGE_IS_5705_OR_BEYOND() to BGE_IS_5705_PLUS() for consistency.
2006-12-04 22:12:21 +00:00
if (!(BGE_IS_5705_PLUS(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
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.
*/
Add support to bge(4) to not break IPMI support when the driver attaches to it. Try to co-operate with the IPMI/ASF firmware accessing the PHY. One we get link we don't mess with the PHY. If we do then over time the NIC will go off line. It would be nice if we could tell if IPMI was enabled on the chip but I can't figure out a reliable way to do that. The scheme I tried worked on a Dell PE850 but not on an HP machine. So we assume any NIC that has ASF capability needs to deal with it. The code was inspired by the support in Linux from kernel.org and Broadcom. Broadcom did give me some info. but it is rather limited and is mostly just what is in the Linux driver. Thanks to the numerous people that helped debug the many prior versions and that I didn't break other bge(4) HW. Reviewed by: several people Tested by: even more
2006-09-09 03:36:57 +00:00
bge_stop_fw(sc);
bge_sig_pre_reset(sc, BGE_RESET_STOP);
bge_reset(sc);
bge_sig_legacy(sc, BGE_RESET_STOP);
bge_sig_post_reset(sc, BGE_RESET_STOP);
/*
* Keep the ASF firmware running if up.
*/
if (sc->bge_asf_mode & ASF_STACKUP)
BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
else
BGE_CLRBIT(sc, BGE_MODE_CTL, BGE_MODECTL_STACKUP);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 the RX lists. */
bge_free_rx_ring_std(sc);
/* Free jumbo RX list. */
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
if (BGE_IS_JUMBO_CAPABLE(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
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.
*/
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if ((sc->bge_flags & BGE_FLAG_TBI) == 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
itmp = ifp->if_flags;
ifp->if_flags |= IFF_UP;
When bge_stop() is called from bge_detach(), mii is already NULL, so check before trying to use it. This fix allows to 'kldunload if_bge' without panicing. Reviewed by: bz MFC after: 3 days
2005-09-04 06:35:59 +00:00
/*
* If we are called from bge_detach(), mii is already NULL.
*/
if (mii != NULL) {
ifm = mii->mii_media.ifm_cur;
mtmp = ifm->ifm_media;
ifm->ifm_media = IFM_ETHER|IFM_NONE;
mii_mediachg(mii);
ifm->ifm_media = mtmp;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
ifp->if_flags = itmp;
}
sc->bge_tx_saved_considx = BGE_TXCONS_UNSET;
1) Ignore link events for MII/GMII cards if MI auto-polling disabled. This should fix strange link state behaviour reported for bcm5721 & bcm5704c 2) Clear bge_link flag in bge_stop() 3) Force link state check after bge_ifmedia_upd(). Otherwise we can miss link event if PHY changes it's state fast enough. Tested by: phk (bcm5704c) Approved by: glebius (mentor) MFC after: 1 week
2006-02-17 14:33:35 +00:00
/*
* We can't just call bge_link_upd() cause chip is almost stopped so
* bge_link_upd -> bge_tick_locked -> bge_stats_update sequence may
* lead to hardware deadlock. So we just clearing MAC's link state
* (PHY may still have link UP).
*/
if (bootverbose && sc->bge_link)
if_printf(sc->bge_ifp, "link DOWN\n");
sc->bge_link = 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
1) Ignore link events for MII/GMII cards if MI auto-polling disabled. This should fix strange link state behaviour reported for bcm5721 & bcm5704c 2) Clear bge_link flag in bge_stop() 3) Force link state check after bge_ifmedia_upd(). Otherwise we can miss link event if PHY changes it's state fast enough. Tested by: phk (bcm5704c) Approved by: glebius (mentor) MFC after: 1 week
2006-02-17 14:33:35 +00:00
ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
/*
* Stop all chip I/O so that the kernel's probe routines don't
* get confused by errant DMAs when rebooting.
*/
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_shutdown(device_t dev)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
{
struct bge_softc *sc;
sc = device_get_softc(dev);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_LOCK(sc);
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
bge_stop(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
bge_reset(sc);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
BGE_UNLOCK(sc);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
Implement suspend/resume methods to be more ACPI friendly. I'm able to suspend/resume my laptop without this change, but then I need to wait for the watchdog to reset the card. With this change, it is ready immediately. Glanced at by: glebius
2005-09-28 19:20:49 +00:00
static int
bge_suspend(device_t dev)
{
struct bge_softc *sc;
sc = device_get_softc(dev);
BGE_LOCK(sc);
bge_stop(sc);
BGE_UNLOCK(sc);
return (0);
}
static int
bge_resume(device_t dev)
{
struct bge_softc *sc;
struct ifnet *ifp;
sc = device_get_softc(dev);
BGE_LOCK(sc);
ifp = sc->bge_ifp;
if (ifp->if_flags & IFF_UP) {
bge_init_locked(sc);
if (ifp->if_drv_flags & IFF_DRV_RUNNING)
bge_start_locked(ifp);
}
BGE_UNLOCK(sc);
return (0);
}
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
static void
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
bge_link_upd(struct bge_softc *sc)
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
{
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
struct mii_data *mii;
uint32_t link, status;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
BGE_LOCK_ASSERT(sc);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Clear 'pending link event' flag. */
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
sc->bge_link_evt = 0;
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
/*
* Process link state changes.
* Grrr. The link status word in the status block does
* not work correctly on the BCM5700 rev AX and BX chips,
* according to all available information. Hence, we have
* to enable MII interrupts in order to properly obtain
* async link changes. Unfortunately, this also means that
* we have to read the MAC status register to detect link
* changes, thereby adding an additional register access to
* the interrupt handler.
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
*
* XXX: perhaps link state detection procedure used for
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
* BGE_CHIPID_BCM5700_B2 can be used for others BCM5700 revisions.
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
*/
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
Possess some work from OpenBSD, with some local additions. - Add more device IDs, ASIC revisions and chip IDs. - Rewrite a bit code that picks the description for device. - Introduce several macros to shorten quirks for bugs and features.[*] - Use some magic values, that OpenBSD has successfully possessed from Linux (Broadcom supplied) driver. - Remove disabled code that tried to access VPD. [*] The macro that matches Jumbo capable NICs is rewritten to preserve our current behavior. I need clarify whether our or theirs is correct. PR: 68351 (and may be others) Obtained from: OpenBSD, brad@ mostly
2006-06-15 14:31:49 +00:00
sc->bge_chipid != BGE_CHIPID_BCM5700_B2) {
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
status = CSR_READ_4(sc, BGE_MAC_STS);
if (status & BGE_MACSTAT_MI_INTERRUPT) {
callout_stop(&sc->bge_stat_ch);
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
bge_tick(sc);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
mii = device_get_softc(sc->bge_miibus);
if (!sc->bge_link &&
mii->mii_media_status & IFM_ACTIVE &&
IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
sc->bge_link++;
if (bootverbose)
if_printf(sc->bge_ifp, "link UP\n");
} else if (sc->bge_link &&
(!(mii->mii_media_status & IFM_ACTIVE) ||
IFM_SUBTYPE(mii->mii_media_active) == IFM_NONE)) {
sc->bge_link = 0;
if (bootverbose)
if_printf(sc->bge_ifp, "link DOWN\n");
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Clear the interrupt. */
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
CSR_WRITE_4(sc, BGE_MAC_EVT_ENB,
BGE_EVTENB_MI_INTERRUPT);
bge_miibus_readreg(sc->bge_dev, 1, BRGPHY_MII_ISR);
bge_miibus_writereg(sc->bge_dev, 1, BRGPHY_MII_IMR,
BRGPHY_INTRS);
}
return;
Whitespace.
2006-06-08 10:19:16 +00:00
}
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_TBI) {
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
status = CSR_READ_4(sc, BGE_MAC_STS);
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
if (status & BGE_MACSTAT_TBI_PCS_SYNCHED) {
if (!sc->bge_link) {
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
sc->bge_link++;
if (sc->bge_asicrev == BGE_ASICREV_BCM5704)
BGE_CLRBIT(sc, BGE_MAC_MODE,
BGE_MACMODE_TBI_SEND_CFGS);
CSR_WRITE_4(sc, BGE_MAC_STS, 0xFFFFFFFF);
if (bootverbose)
if_printf(sc->bge_ifp, "link UP\n");
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
if_link_state_change(sc->bge_ifp,
LINK_STATE_UP);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
}
1) Fix link state detection for cards operating in TBI mode (fiber ones) 2) add missing bus_dmamap_sync() call in bge_intr() Tested by: Husnu Demir <hdemir AT metu DOT edu DOT tr> Approved by: glebius (mentor) MFC after: 3 days
2006-02-01 14:26:35 +00:00
} else if (sc->bge_link) {
sc->bge_link = 0;
if (bootverbose)
if_printf(sc->bge_ifp, "link DOWN\n");
if_link_state_change(sc->bge_ifp, LINK_STATE_DOWN);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
}
1) Ignore link events for MII/GMII cards if MI auto-polling disabled. This should fix strange link state behaviour reported for bcm5721 & bcm5704c 2) Clear bge_link flag in bge_stop() 3) Force link state check after bge_ifmedia_upd(). Otherwise we can miss link event if PHY changes it's state fast enough. Tested by: phk (bcm5704c) Approved by: glebius (mentor) MFC after: 1 week
2006-02-17 14:33:35 +00:00
/* Discard link events for MII/GMII cards if MI auto-polling disabled */
} else if (CSR_READ_4(sc, BGE_MI_MODE) & BGE_MIMODE_AUTOPOLL) {
Whitespace.
2006-06-08 10:19:16 +00:00
/*
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
* Some broken BCM chips have BGE_STATFLAG_LINKSTATE_CHANGED bit
* in status word always set. Workaround this bug by reading
* PHY link status directly.
*/
link = (CSR_READ_4(sc, BGE_MI_STS) & BGE_MISTS_LINK) ? 1 : 0;
if (link != sc->bge_link ||
sc->bge_asicrev == BGE_ASICREV_BCM5700) {
callout_stop(&sc->bge_stat_ch);
- Instead of if_watchdog/if_timer interface use our own timer that piggybacks on bge_tick() callout. - Lock bge_tick() using callout_init_mtx().
2006-11-30 13:40:39 +00:00
bge_tick(sc);
1) move all link state detection code from bge_tick_locked() to bge_link_upd() 2) use more robust way of link state handling for BCM5700 rev.B2 chip 3) workaround bug of some BCM570x chips which cause spurious "link up" messages 4) fix bug: some BCM570x chips was unable to detect link state changes after ifconfig down/up sequence until any 'non-link related' interrupt generated. (this happened due to pending internal link state attention which blocked interrupt generation) Approved by: glebius (mentor) MFC after: 1 week
2006-01-13 08:59:40 +00:00
mii = device_get_softc(sc->bge_miibus);
if (!sc->bge_link &&
mii->mii_media_status & IFM_ACTIVE &&
IFM_SUBTYPE(mii->mii_media_active) != IFM_NONE) {
sc->bge_link++;
if (bootverbose)
if_printf(sc->bge_ifp, "link UP\n");
} else if (sc->bge_link &&
(!(mii->mii_media_status & IFM_ACTIVE) ||
IFM_SUBTYPE(mii->mii_media_active) == IFM_NONE)) {
sc->bge_link = 0;
if (bootverbose)
if_printf(sc->bge_ifp, "link DOWN\n");
}
}
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Clear the attention. */
1) fix tiny bug in bge_start_locked() 2) rework link state detection code & use it in POLLING mode 3) fix 2 bugs in link state detection code: a) driver unable to detect link loss on bcm5721 b) on bcm570x chips (tested on bcm5700 bcm5701 bcm5702) driver fails to detect link loss with probability 1/6 (solved in brgphy.c) Devices working in TBI mode should not be affected by this change. Approved by: glebius (mentor) MFC after: 1 month
2005-12-08 13:31:52 +00:00
CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED|
BGE_MACSTAT_CFG_CHANGED|BGE_MACSTAT_MI_COMPLETE|
BGE_MACSTAT_LINK_CHANGED);
}
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
static void
bge_add_sysctls(struct bge_softc *sc)
{
struct sysctl_ctx_list *ctx;
struct sysctl_oid_list *children;
ctx = device_get_sysctl_ctx(sc->bge_dev);
children = SYSCTL_CHILDREN(device_get_sysctl_tree(sc->bge_dev));
#ifdef BGE_REGISTER_DEBUG
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "debug_info",
CTLTYPE_INT | CTLFLAG_RW, sc, 0, bge_sysctl_debug_info, "I",
"Debug Information");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "reg_read",
CTLTYPE_INT | CTLFLAG_RW, sc, 0, bge_sysctl_reg_read, "I",
"Register Read");
SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "mem_read",
CTLTYPE_INT | CTLFLAG_RW, sc, 0, bge_sysctl_mem_read, "I",
"Memory Read");
SYSCTL_ADD_ULONG(ctx, children, OID_AUTO, "stat_IfHcInOctets",
CTLFLAG_RD,
&sc->bge_ldata.bge_stats->rxstats.ifHCInOctets.bge_addr_lo,
"Bytes received");
SYSCTL_ADD_ULONG(ctx, children, OID_AUTO, "stat_IfHcOutOctets",
CTLFLAG_RD,
&sc->bge_ldata.bge_stats->txstats.ifHCOutOctets.bge_addr_lo,
"Bytes received");
#endif
}
#ifdef BGE_REGISTER_DEBUG
static int
bge_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
{
struct bge_softc *sc;
uint16_t *sbdata;
int error;
int result;
int i, j;
result = -1;
error = sysctl_handle_int(oidp, &result, 0, req);
if (error || (req->newptr == NULL))
return (error);
if (result == 1) {
sc = (struct bge_softc *)arg1;
sbdata = (uint16_t *)sc->bge_ldata.bge_status_block;
printf("Status Block:\n");
for (i = 0x0; i < (BGE_STATUS_BLK_SZ / 4); ) {
printf("%06x:", i);
for (j = 0; j < 8; j++) {
printf(" %04x", sbdata[i]);
i += 4;
}
printf("\n");
}
printf("Registers:\n");
for (i = 0x800; i < 0xa00; ) {
printf("%06x:", i);
for (j = 0; j < 8; j++) {
printf(" %08x", CSR_READ_4(sc, i));
i += 4;
}
printf("\n");
}
printf("Hardware Flags:\n");
Use the BGE_IS_* macros consistently. Also add a couple of missing cases to the debug_info function.
2006-12-13 21:03:55 +00:00
if (BGE_IS_575X_PLUS(sc))
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
printf(" - 575X Plus\n");
Use the BGE_IS_* macros consistently. Also add a couple of missing cases to the debug_info function.
2006-12-13 21:03:55 +00:00
if (BGE_IS_5705_PLUS(sc))
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
printf(" - 5705 Plus\n");
Use the BGE_IS_* macros consistently. Also add a couple of missing cases to the debug_info function.
2006-12-13 21:03:55 +00:00
if (BGE_IS_5714_FAMILY(sc))
printf(" - 5714 Family\n");
if (BGE_IS_5700_FAMILY(sc))
printf(" - 5700 Family\n");
Fix support for certain 575x/578x chips. This consists of the following: - Use the appropriate register writing method when reseting the chip - Program the descriptor DMA engine correctly. - More reliably detect certain chips and their features. Also add some low-level debugging tools to help future work on this driver. Submitted by: David Christenson (proof of concept changes) Sponsored by: www.UIA.net
2006-12-12 05:11:12 +00:00
if (sc->bge_flags & BGE_FLAG_JUMBO)
printf(" - Supports Jumbo Frames\n");
if (sc->bge_flags & BGE_FLAG_PCIX)
printf(" - PCI-X Bus\n");
if (sc->bge_flags & BGE_FLAG_PCIE)
printf(" - PCI Express Bus\n");
if (sc->bge_flags & BGE_FLAG_NO3LED)
printf(" - No 3 LEDs\n");
if (sc->bge_flags & BGE_FLAG_RX_ALIGNBUG)
printf(" - RX Alignment Bug\n");
}
return (error);
}
static int
bge_sysctl_reg_read(SYSCTL_HANDLER_ARGS)
{
struct bge_softc *sc;
int error;
uint16_t result;
uint32_t val;
result = -1;
error = sysctl_handle_int(oidp, &result, 0, req);
if (error || (req->newptr == NULL))
return (error);
if (result < 0x8000) {
sc = (struct bge_softc *)arg1;
val = CSR_READ_4(sc, result);
printf("reg 0x%06X = 0x%08X\n", result, val);
}
return (error);
}
static int
bge_sysctl_mem_read(SYSCTL_HANDLER_ARGS)
{
struct bge_softc *sc;
int error;
uint16_t result;
uint32_t val;
result = -1;
error = sysctl_handle_int(oidp, &result, 0, req);
if (error || (req->newptr == NULL))
return (error);
if (result < 0x8000) {
sc = (struct bge_softc *)arg1;
val = bge_readmem_ind(sc, result);
printf("mem 0x%06X = 0x%08X\n", result, val);
}
return (error);
}
#endif
Reference in New Issue Copy Permalink