d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
33e98ade53
freebsd-nq/sys/dev/bge/if_bge.c

5202 lines
149 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>
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
#include <sys/taskqueue.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>
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
#include <netinet/tcp.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 <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>
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
#ifdef __sparc64__
#include <dev/ofw/ofw_bus.h>
#include <dev/ofw/openfirm.h>
#include <machine/ofw_machdep.h>
#include <machine/ver.h>
#endif
Prefer new location of pci include files (which have only been in the tree for two or more years now), except in a few places where there's code to be compatible with older versions of FreeBSD.
2003-08-22 05:54:52 +00:00
#include <dev/pci/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>
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
#define BGE_CSUM_FEATURES (CSUM_IP | CSUM_TCP)
Fix style(9). Pointed out by: many
2007-02-12 23:33:05 +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.
*/
- Don't read the identifier string from the VPD if there's no chip containing an Ethernet address fitted as this is yet another thing that fails in that case in order to avoid the one second delay until pci_read_vpd_reg() times out. - Const'ify the bge_devs array.
2008-09-22 20:55:19 +00:00
static const struct bge_type {
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
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 },
Add support for the BMC5722. Reported by: Chris Shenton <chris at shenton dot org> on current@ Approved by: imp (mentor)
2007-12-25 19:51:18 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5722 },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5723 },
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_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 },
o Add PCI ID for BCM 5756. o Don't enable BGE_FLAG_BER_BUG on both 5722 and 5756, and based on their PCI IDs rather than their chip IDs. Reported by: several PC-BSD users via kmoore Reviewed by: yongari, imp, jhb, davidch Sponsored by: iXsystems, Inc. MFC after: 2 weeks
2010-01-13 22:39:39 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5756 },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5761 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5761E },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5761S },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5761SE },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5764 },
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 },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5784 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5785F },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5785G },
- 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 },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5787F },
- 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_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 },
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5906 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM5906M },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57760 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57780 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57788 },
{ BCOM_VENDORID, BCOM_DEVICEID_BCM57790 },
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
{ SK_VENDORID, SK_DEVICEID_ALTIMA },
{ TC_VENDORID, TC_DEVICEID_3C996 },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ FJTSU_VENDORID, FJTSU_DEVICEID_PW008GE4 },
{ FJTSU_VENDORID, FJTSU_DEVICEID_PW008GE5 },
{ FJTSU_VENDORID, FJTSU_DEVICEID_PP250450 },
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, 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" },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ FJTSU_VENDORID, "Fujitsu" },
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 }
};
Remove extra white space.
2009-11-22 18:30:19 +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
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" },
Add more chipset revision IDs. Obtained from: OpenBSD
2007-03-09 01:30:23 +00:00
{ BGE_CHIPID_BCM5715_A3, "BCM5715 A3" },
{ BGE_CHIPID_BCM5755_A0, "BCM5755 A0" },
{ BGE_CHIPID_BCM5755_A1, "BCM5755 A1" },
{ BGE_CHIPID_BCM5755_A2, "BCM5755 A2" },
Flesh out support for the BCM5722 by recognizing the phy on the 5722 and the specific ASIC revision. MFC after: 1 week Obtained from: OpenBSD (mii/phy bits)
2008-03-06 21:42:48 +00:00
{ BGE_CHIPID_BCM5722_A0, "BCM5722 A0" },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BGE_CHIPID_BCM5761_A0, "BCM5761 A0" },
{ BGE_CHIPID_BCM5761_A1, "BCM5761 A1" },
{ BGE_CHIPID_BCM5784_A0, "BCM5784 A0" },
{ BGE_CHIPID_BCM5784_A1, "BCM5784 A1" },
Fix two typos in comments.
2007-02-14 19:01:33 +00:00
/* 5754 and 5787 share the same ASIC ID */
Remove extra white space.
2009-11-22 18:30:19 +00:00
{ BGE_CHIPID_BCM5787_A0, "BCM5754/5787 A0" },
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_CHIPID_BCM5787_A1, "BCM5754/5787 A1" },
{ BGE_CHIPID_BCM5787_A2, "BCM5754/5787 A2" },
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
{ BGE_CHIPID_BCM5906_A1, "BCM5906 A1" },
{ BGE_CHIPID_BCM5906_A2, "BCM5906 A2" },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BGE_CHIPID_BCM57780_A0, "BCM57780 A0" },
{ BGE_CHIPID_BCM57780_A1, "BCM57780 A1" },
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" },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BGE_ASICREV_BCM5761, "unknown BCM5761" },
{ BGE_ASICREV_BCM5784, "unknown BCM5784" },
{ BGE_ASICREV_BCM5785, "unknown BCM5785" },
Fix two typos in comments.
2007-02-14 19:01:33 +00:00
/* 5754 and 5787 share the same ASIC ID */
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_ASICREV_BCM5787, "unknown BCM5754/5787" },
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
{ BGE_ASICREV_BCM5906, "unknown BCM5906" },
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
{ BGE_ASICREV_BCM57780, "unknown BCM57780" },
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
};
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +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)
#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)
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
#define BGE_IS_5755_PLUS(sc) ((sc)->bge_flags & BGE_FLAG_5755_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);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
typedef int (*bge_eaddr_fcn_t)(struct bge_softc *, uint8_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);
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
static int bge_dma_alloc(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_dma_free(struct bge_softc *);
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
static int bge_dma_ring_alloc(struct bge_softc *, bus_size_t, bus_size_t,
bus_dma_tag_t *, uint8_t **, bus_dmamap_t *, bus_addr_t *, const char *);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
static int bge_get_eaddr_fw(struct bge_softc *sc, uint8_t ether_addr[]);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
static int bge_get_eaddr_mem(struct bge_softc *, uint8_t[]);
static int bge_get_eaddr_nvram(struct bge_softc *, uint8_t[]);
static int bge_get_eaddr_eeprom(struct bge_softc *, uint8_t[]);
static int bge_get_eaddr(struct bge_softc *, uint8_t[]);
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
static void bge_txeof(struct bge_softc *, uint16_t);
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
static int bge_rxeof(struct bge_softc *, uint16_t, int);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +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 *);
- 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 *);
Add missing function prototype in r199671.
2009-11-22 21:20:26 +00:00
static struct mbuf *bge_setup_tso(struct bge_softc *, struct mbuf *,
uint16_t *);
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 *);
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
static int bge_msi_intr(void *);
static void bge_intr_task(void *, int);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
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 *);
device_shutdown returns an int.
2009-02-05 18:43:13 +00:00
static int 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 *);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
static uint8_t bge_nvram_getbyte(struct bge_softc *, int, uint8_t *);
static int bge_read_nvram(struct bge_softc *, caddr_t, int, int);
- 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 *, 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 *);
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
static void bge_setvlan(struct bge_softc *);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
static __inline void bge_rxreuse_std(struct bge_softc *, int);
static __inline void bge_rxreuse_jumbo(struct bge_softc *, int);
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
static int bge_newbuf_std(struct bge_softc *, int);
static int bge_newbuf_jumbo(struct bge_softc *, int);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
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
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
static int bge_has_eaddr(struct bge_softc *);
- 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 support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
static void bge_writembx(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
Fix recent style problems and expand a comment. Submitted by: Bruce Evans
2006-12-13 21:13:09 +00:00
static void bge_writemem_direct(struct bge_softc *, int, int);
- 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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int 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
Fix style(9). Pointed out by: many
2007-02-12 23:33:05 +00:00
#define BGE_RESET_START 1
#define BGE_RESET_STOP 2
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_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);
Make bge_stop_fw() static. While I'm here use ANSI function definitions.
2010-07-13 19:33:46 +00:00
static void bge_stop_fw(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 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 MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
static int bge_sysctl_stats(SYSCTL_HANDLER_ARGS);
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
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
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;
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, 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
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
#define SPARC64_BLADE_1500_MODEL "SUNW,Sun-Blade-1500"
#define SPARC64_BLADE_1500_PATH_BGE "/pci@1f,700000/network@2"
#define SPARC64_BLADE_2500_MODEL "SUNW,Sun-Blade-2500"
#define SPARC64_BLADE_2500_PATH_BGE "/pci@1c,600000/network@3"
#define SPARC64_OFW_SUBVENDOR "subsystem-vendor-id"
static int
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
bge_has_eaddr(struct bge_softc *sc)
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
{
#ifdef __sparc64__
char buf[sizeof(SPARC64_BLADE_1500_PATH_BGE)];
device_t dev;
uint32_t subvendor;
dev = sc->bge_dev;
/*
* The on-board BGEs found in sun4u machines aren't fitted with
* an EEPROM which means that we have to obtain the MAC address
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
* via OFW and that some tests will always fail. We distinguish
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
* such BGEs by the subvendor ID, which also has to be obtained
* from OFW instead of the PCI configuration space as the latter
* indicates Broadcom as the subvendor of the netboot interface.
* For early Blade 1500 and 2500 we even have to check the OFW
* device path as the subvendor ID always defaults to Broadcom
* there.
*/
if (OF_getprop(ofw_bus_get_node(dev), SPARC64_OFW_SUBVENDOR,
&subvendor, sizeof(subvendor)) == sizeof(subvendor) &&
For the on-board interfaces found in Fujitsu SPARC64 machines obtain the MAC address via OFW as well.
2010-04-26 18:56:06 +00:00
(subvendor == FJTSU_VENDORID || subvendor == SUN_VENDORID))
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
return (0);
memset(buf, 0, sizeof(buf));
if (OF_package_to_path(ofw_bus_get_node(dev), buf, sizeof(buf)) > 0) {
if (strcmp(sparc64_model, SPARC64_BLADE_1500_MODEL) == 0 &&
strcmp(buf, SPARC64_BLADE_1500_PATH_BGE) == 0)
return (0);
if (strcmp(sparc64_model, SPARC64_BLADE_2500_MODEL) == 0 &&
strcmp(buf, SPARC64_BLADE_2500_PATH_BGE) == 0)
return (0);
}
#endif
return (1);
}
- 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);
}
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
static void
bge_writembx(struct bge_softc *sc, int off, int val)
{
if (sc->bge_asicrev == BGE_ASICREV_BCM5906)
off += BGE_LPMBX_IRQ0_HI - BGE_MBX_IRQ0_HI;
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;
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
KASSERT(nseg == 1, ("%s: %d segments returned!", __func__, nseg));
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
ctx = arg;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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_busaddr = segs->ds_addr;
}
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
static uint8_t
bge_nvram_getbyte(struct bge_softc *sc, int addr, uint8_t *dest)
{
uint32_t access, byte = 0;
int i;
/* Lock. */
CSR_WRITE_4(sc, BGE_NVRAM_SWARB, BGE_NVRAMSWARB_SET1);
for (i = 0; i < 8000; i++) {
if (CSR_READ_4(sc, BGE_NVRAM_SWARB) & BGE_NVRAMSWARB_GNT1)
break;
DELAY(20);
}
if (i == 8000)
return (1);
/* Enable access. */
access = CSR_READ_4(sc, BGE_NVRAM_ACCESS);
CSR_WRITE_4(sc, BGE_NVRAM_ACCESS, access | BGE_NVRAMACC_ENABLE);
CSR_WRITE_4(sc, BGE_NVRAM_ADDR, addr & 0xfffffffc);
CSR_WRITE_4(sc, BGE_NVRAM_CMD, BGE_NVRAM_READCMD);
for (i = 0; i < BGE_TIMEOUT * 10; i++) {
DELAY(10);
if (CSR_READ_4(sc, BGE_NVRAM_CMD) & BGE_NVRAMCMD_DONE) {
DELAY(10);
break;
}
}
if (i == BGE_TIMEOUT * 10) {
if_printf(sc->bge_ifp, "nvram read timed out\n");
return (1);
}
/* Get result. */
byte = CSR_READ_4(sc, BGE_NVRAM_RDDATA);
*dest = (bswap32(byte) >> ((addr % 4) * 8)) & 0xFF;
/* Disable access. */
CSR_WRITE_4(sc, BGE_NVRAM_ACCESS, access);
/* Unlock. */
CSR_WRITE_4(sc, BGE_NVRAM_SWARB, BGE_NVRAMSWARB_CLR1);
CSR_READ_4(sc, BGE_NVRAM_SWARB);
return (0);
}
/*
* Read a sequence of bytes from NVRAM.
*/
static int
bge_read_nvram(struct bge_softc *sc, caddr_t dest, int off, int cnt)
{
int err = 0, i;
uint8_t byte = 0;
if (sc->bge_asicrev != BGE_ASICREV_BCM5906)
return (1);
for (i = 0; i < cnt; i++) {
err = bge_nvram_getbyte(sc, off + i, &byte);
if (err)
break;
*(dest + i) = byte;
}
return (err ? 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
/*
* 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,
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
BGE_EEADDR_RESET | BGE_EEHALFCLK(BGE_HALFCLK_384SCL));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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(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;
}
Fix time out check in EEPROM read and move delays to give some settle time. Submitted by: mjacob
2007-05-22 18:51:05 +00:00
if (i == BGE_TIMEOUT * 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, "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);
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +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);
}
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
CSR_WRITE_4(sc, BGE_MI_COMM, BGE_MICMD_READ | BGE_MICOMM_BUSY |
BGE_MIPHY(phy) | BGE_MIREG(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
for (i = 0; i < BGE_TIMEOUT; i++) {
Fix time out check in EEPROM read and move delays to give some settle time. Submitted by: mjacob
2007-05-22 18:51:05 +00:00
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
val = CSR_READ_4(sc, BGE_MI_COMM);
if (!(val & BGE_MICOMM_BUSY))
break;
}
if (i == BGE_TIMEOUT) {
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
device_printf(sc->bge_dev,
"PHY read timed out (phy %d, reg %d, val 0x%08x)\n",
phy, reg, val);
- 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
}
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
DELAY(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
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
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +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);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5906 &&
(reg == BRGPHY_MII_1000CTL || reg == BRGPHY_MII_AUXCTL))
style.
2010-07-13 19:39:51 +00:00
return (0);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +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);
}
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
CSR_WRITE_4(sc, BGE_MI_COMM, BGE_MICMD_WRITE | BGE_MICOMM_BUSY |
BGE_MIPHY(phy) | BGE_MIREG(reg) | 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
for (i = 0; i < BGE_TIMEOUT; i++) {
Fix time out check in EEPROM read and move delays to give some settle time. Submitted by: mjacob
2007-05-22 18:51:05 +00:00
DELAY(10);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
if (!(CSR_READ_4(sc, BGE_MI_COMM) & BGE_MICOMM_BUSY)) {
DELAY(5);
CSR_READ_4(sc, BGE_MI_COMM); /* dummy read */
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
}
Fix time out check in EEPROM read and move delays to give some settle time. Submitted by: mjacob
2007-05-22 18:51:05 +00:00
}
if (i == BGE_TIMEOUT) {
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
device_printf(sc->bge_dev,
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
"PHY write timed out (phy %d, reg %d, val %d)\n",
phy, reg, val);
Fix time out check in EEPROM read and move delays to give some settle time. Submitted by: mjacob
2007-05-22 18:51:05 +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
}
- 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);
}
- 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);
For controllers that has dual mode PHY(copper or fiber) interfaces over GMII, make sure to enable GMII. With this change brgphy(4) is used to handle the dual mode PHY. Since we still don't have a sane way to pass PHY specific information to mii(4) layer special handling is needed in brgphy(4) to determine which mode of PHY was configured in parent interface. This change make BCM5715S work. Tested by: olli Obtained from: OpenBSD MFC after: 1 week
2010-01-14 19:08:43 +00:00
if (IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_T ||
IFM_SUBTYPE(mii->mii_media_active) == IFM_1000_SX)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
bge_newbuf_std(struct bge_softc *sc, 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
{
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
struct mbuf *m;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
struct bge_rx_bd *r;
Covert bge_newbuf_std to use bus_dmamap_load_mbuf_sg(9). Note, bge_newbuf_std still has a bug for handling dma map load failure under high network load. Just reusing mbuf is not enough as driver already unloaded the dma map of the mbuf. Graceful recovery needs more work. Ideally we can just update dma address part of a Rx descriptor because the controller never overwrite the Rx descriptor. This requires some Rx initialization code changes and it would be done later after fixing other incorrect bus_dma(9) usages.
2009-11-04 20:40:38 +00:00
bus_dma_segment_t segs[1];
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
bus_dmamap_t map;
Covert bge_newbuf_std to use bus_dmamap_load_mbuf_sg(9). Note, bge_newbuf_std still has a bug for handling dma map load failure under high network load. Just reusing mbuf is not enough as driver already unloaded the dma map of the mbuf. Graceful recovery needs more work. Ideally we can just update dma address part of a Rx descriptor because the controller never overwrite the Rx descriptor. This requires some Rx initialization code changes and it would be done later after fixing other incorrect bus_dma(9) usages.
2009-11-04 20:40:38 +00:00
int error, 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
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
if (m == NULL)
return (ENOBUFS);
m->m_len = m->m_pkthdr.len = MCLBYTES;
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)
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
m_adj(m, ETHER_ALIGN);
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_rx_mtag,
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
sc->bge_cdata.bge_rx_std_sparemap, m, segs, &nsegs, 0);
Covert bge_newbuf_std to use bus_dmamap_load_mbuf_sg(9). Note, bge_newbuf_std still has a bug for handling dma map load failure under high network load. Just reusing mbuf is not enough as driver already unloaded the dma map of the mbuf. Graceful recovery needs more work. Ideally we can just update dma address part of a Rx descriptor because the controller never overwrite the Rx descriptor. This requires some Rx initialization code changes and it would be done later after fixing other incorrect bus_dma(9) usages.
2009-11-04 20:40:38 +00:00
if (error != 0) {
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
m_freem(m);
Covert bge_newbuf_std to use bus_dmamap_load_mbuf_sg(9). Note, bge_newbuf_std still has a bug for handling dma map load failure under high network load. Just reusing mbuf is not enough as driver already unloaded the dma map of the mbuf. Graceful recovery needs more work. Ideally we can just update dma address part of a Rx descriptor because the controller never overwrite the Rx descriptor. This requires some Rx initialization code changes and it would be done later after fixing other incorrect bus_dma(9) usages.
2009-11-04 20:40:38 +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
}
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
if (sc->bge_cdata.bge_rx_std_chain[i] != NULL) {
bus_dmamap_sync(sc->bge_cdata.bge_rx_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i], BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->bge_cdata.bge_rx_mtag,
sc->bge_cdata.bge_rx_std_dmamap[i]);
}
map = sc->bge_cdata.bge_rx_std_dmamap[i];
sc->bge_cdata.bge_rx_std_dmamap[i] = sc->bge_cdata.bge_rx_std_sparemap;
sc->bge_cdata.bge_rx_std_sparemap = map;
sc->bge_cdata.bge_rx_std_chain[i] = m;
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
sc->bge_cdata.bge_rx_std_seglen[i] = segs[0].ds_len;
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
r = &sc->bge_ldata.bge_rx_std_ring[sc->bge_std];
Covert bge_newbuf_std to use bus_dmamap_load_mbuf_sg(9). Note, bge_newbuf_std still has a bug for handling dma map load failure under high network load. Just reusing mbuf is not enough as driver already unloaded the dma map of the mbuf. Graceful recovery needs more work. Ideally we can just update dma address part of a Rx descriptor because the controller never overwrite the Rx descriptor. This requires some Rx initialization code changes and it would be done later after fixing other incorrect bus_dma(9) usages.
2009-11-04 20:40:38 +00:00
r->bge_addr.bge_addr_lo = BGE_ADDR_LO(segs[0].ds_addr);
r->bge_addr.bge_addr_hi = BGE_ADDR_HI(segs[0].ds_addr);
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_END;
Covert bge_newbuf_std to use bus_dmamap_load_mbuf_sg(9). Note, bge_newbuf_std still has a bug for handling dma map load failure under high network load. Just reusing mbuf is not enough as driver already unloaded the dma map of the mbuf. Graceful recovery needs more work. Ideally we can just update dma address part of a Rx descriptor because the controller never overwrite the Rx descriptor. This requires some Rx initialization code changes and it would be done later after fixing other incorrect bus_dma(9) usages.
2009-11-04 20:40:38 +00:00
r->bge_len = segs[0].ds_len;
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_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
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_mtag,
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
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
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
bge_newbuf_jumbo(struct bge_softc *sc, 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
{
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];
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
bus_dmamap_t map;
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
struct bge_extrx_bd *r;
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
struct mbuf *m;
int error, nsegs;
MGETHDR(m, M_DONTWAIT, MT_DATA);
if (m == NULL)
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
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
m_cljget(m, M_DONTWAIT, MJUM9BYTES);
if (!(m->m_flags & M_EXT)) {
m_freem(m);
return (ENOBUFS);
}
m->m_len = m->m_pkthdr.len = MJUM9BYTES;
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)
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
m_adj(m, 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,
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
sc->bge_cdata.bge_rx_jumbo_sparemap, m, segs, &nsegs, 0);
if (error != 0) {
m_freem(m);
- 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
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
if (sc->bge_cdata.bge_rx_jumbo_chain[i] == NULL) {
bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i], BUS_DMASYNC_POSTREAD);
bus_dmamap_unload(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_dmamap[i]);
}
map = sc->bge_cdata.bge_rx_jumbo_dmamap[i];
sc->bge_cdata.bge_rx_jumbo_dmamap[i] =
sc->bge_cdata.bge_rx_jumbo_sparemap;
sc->bge_cdata.bge_rx_jumbo_sparemap = map;
sc->bge_cdata.bge_rx_jumbo_chain[i] = m;
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
sc->bge_cdata.bge_rx_jumbo_seglen[i][0] = 0;
sc->bge_cdata.bge_rx_jumbo_seglen[i][1] = 0;
sc->bge_cdata.bge_rx_jumbo_seglen[i][2] = 0;
sc->bge_cdata.bge_rx_jumbo_seglen[i][3] = 0;
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
/*
* Fill in the extended RX buffer descriptor.
*/
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
r = &sc->bge_ldata.bge_rx_jumbo_ring[sc->bge_jumbo];
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
r->bge_flags = BGE_RXBDFLAG_JUMBO_RING | BGE_RXBDFLAG_END;
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_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;
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
sc->bge_cdata.bge_rx_jumbo_seglen[i][3] = segs[3].ds_len;
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
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;
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
sc->bge_cdata.bge_rx_jumbo_seglen[i][2] = segs[2].ds_len;
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
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;
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
sc->bge_cdata.bge_rx_jumbo_seglen[i][1] = segs[1].ds_len;
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
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;
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
sc->bge_cdata.bge_rx_jumbo_seglen[i][0] = segs[0].ds_len;
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
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
Use correct dma tag for jumbo buffer.
2009-11-04 20:19:21 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_mtag_jumbo,
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
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
}
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
{
Make bge_newbuf_std()/bge_newbuf_jumbo() returns actual error code for buffer allocation. If driver know we are out of Rx buffers let controller stop. This should fix panic when interface is run even if it had no configured Rx buffers.
2009-11-04 21:06:54 +00:00
int error, 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
Zero out Tx/Rx descriptors before using them. Also add missing bus_dmamap_sync(9) after Tx descriptor initialization.
2009-11-09 23:09:18 +00:00
bzero(sc->bge_ldata.bge_rx_std_ring, BGE_STD_RX_RING_SZ);
Fix I mssied in r199011. Rx ring index also should be updated. If we fill Rx ring full instead of half we can simplify this logic but this requires more experimentation.
2009-11-07 02:10:59 +00:00
sc->bge_std = 0;
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
if ((error = bge_newbuf_std(sc, i)) != 0)
Make bge_newbuf_std()/bge_newbuf_jumbo() returns actual error code for buffer allocation. If driver know we are out of Rx buffers let controller stop. This should fix panic when interface is run even if it had no configured Rx buffers.
2009-11-04 21:06:54 +00:00
return (error);
Fix I mssied in r199011. Rx ring index also should be updated. If we fill Rx ring full instead of half we can simplify this logic but this requires more experimentation.
2009-11-07 02:10:59 +00:00
BGE_INC(sc->bge_std, BGE_STD_RX_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
};
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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,
Controller does not write Rx descriptors, remove BUS_DMASYNC_PREREAD.
2009-11-22 19:17:32 +00:00
sc->bge_cdata.bge_rx_std_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
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
sc->bge_std = 0;
bge_writembx(sc, BGE_MBX_RX_STD_PROD_LO, BGE_STD_RX_RING_CNT - 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
- 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) {
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_mtag,
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_std_dmamap[i],
BUS_DMASYNC_POSTREAD);
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_rx_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
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;
Make bge_newbuf_std()/bge_newbuf_jumbo() returns actual error code for buffer allocation. If driver know we are out of Rx buffers let controller stop. This should fix panic when interface is run even if it had no configured Rx buffers.
2009-11-04 21:06:54 +00:00
int error, 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
Zero out Tx/Rx descriptors before using them. Also add missing bus_dmamap_sync(9) after Tx descriptor initialization.
2009-11-09 23:09:18 +00:00
bzero(sc->bge_ldata.bge_rx_jumbo_ring, BGE_JUMBO_RX_RING_SZ);
Fix I mssied in r199011. Rx ring index also should be updated. If we fill Rx ring full instead of half we can simplify this logic but this requires more experimentation.
2009-11-07 02:10:59 +00:00
sc->bge_jumbo = 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 < BGE_JUMBO_RX_RING_CNT; i++) {
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
if ((error = bge_newbuf_jumbo(sc, i)) != 0)
Make bge_newbuf_std()/bge_newbuf_jumbo() returns actual error code for buffer allocation. If driver know we are out of Rx buffers let controller stop. This should fix panic when interface is run even if it had no configured Rx buffers.
2009-11-04 21:06:54 +00:00
return (error);
Fix I mssied in r199011. Rx ring index also should be updated. If we fill Rx ring full instead of half we can simplify this logic but this requires more experimentation.
2009-11-07 02:10:59 +00:00
BGE_INC(sc->bge_jumbo, BGE_JUMBO_RX_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
};
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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,
Controller does not write Rx descriptors, remove BUS_DMASYNC_PREREAD.
2009-11-22 19:17:32 +00:00
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
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
sc->bge_jumbo = 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
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
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
bge_writembx(sc, BGE_MBX_RX_JUMBO_PROD_LO, BGE_JUMBO_RX_RING_CNT - 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
- 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) {
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_tx_mtag,
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_tx_dmamap[i],
BUS_DMASYNC_POSTWRITE);
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_tx_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
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
Zero out Tx/Rx descriptors before using them. Also add missing bus_dmamap_sync(9) after Tx descriptor initialization.
2009-11-09 23:09:18 +00:00
bzero(sc->bge_ldata.bge_tx_ring, BGE_TX_RING_SZ);
bus_dmamap_sync(sc->bge_cdata.bge_tx_ring_tag,
Controller does not update Tx descriptors(send BDs) after sending frames so remove unnecessary BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTREAD of bus_dmamap_sync(9).
2009-11-10 20:29:20 +00:00
sc->bge_cdata.bge_tx_ring_map, BUS_DMASYNC_PREWRITE);
Zero out Tx/Rx descriptors before using them. Also add missing bus_dmamap_sync(9) after Tx descriptor initialization.
2009-11-09 23:09:18 +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;
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_TX_HOST_PROD0_LO, sc->bge_tx_prodidx);
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
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)
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(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 support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(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)
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(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;
Partially back out rev. 1.148 and use new ETHER_BPF_MTAP() macro for VLAN. Remaining changes are cosmetic.
2006-12-20 01:12:07 +00:00
/* Enable or disable promiscuous mode as needed. */
Do not strip VLAN tag in promiscuous mode.
2006-09-18 22:18:21 +00:00
if (ifp->if_flags & IFF_PROMISC)
Partially back out rev. 1.148 and use new ETHER_BPF_MTAP() macro for VLAN. Remaining changes are cosmetic.
2006-12-20 01:12:07 +00:00
BGE_SETBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_PROMISC);
Do not strip VLAN tag in promiscuous mode.
2006-09-18 22:18:21 +00:00
else
Partially back out rev. 1.148 and use new ETHER_BPF_MTAP() macro for VLAN. Remaining changes are cosmetic.
2006-12-20 01:12:07 +00:00
BGE_CLRBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_PROMISC);
Do not strip VLAN tag in promiscuous mode.
2006-09-18 22:18:21 +00:00
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
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++)
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
CSR_WRITE_4(sc, BGE_MAR0 + (i * 4), 0xFFFFFFFF);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
}
/* 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. */
Use if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_rlock(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 *)
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
ifma->ifma_addr), ETHER_ADDR_LEN) & 0x7F;
hashes[(h & 0x60) >> 5] |= 1 << (h & 0x1F);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 if_maddr_rlock()/if_maddr_runlock() rather than IF_ADDR_LOCK()/ IF_ADDR_UNLOCK() across network device drivers when accessing the per-interface multicast address list, if_multiaddrs. This will allow us to change the locking strategy without affecting our driver programming interface or binary interface. For two wireless drivers, remove unnecessary locking, since they don't actually access the multicast address list. Approved by: re (kib) MFC after: 6 weeks
2009-06-26 11:45:06 +00:00
if_maddr_runlock(ifp);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
static void
bge_setvlan(struct bge_softc *sc)
{
struct ifnet *ifp;
BGE_LOCK_ASSERT(sc);
ifp = sc->bge_ifp;
/* Enable or disable VLAN tag stripping as needed. */
if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
BGE_CLRBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_KEEP_VLAN_DIAG);
else
BGE_SETBIT(sc, BGE_RX_MODE, BGE_RXMODE_RX_KEEP_VLAN_DIAG);
}
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
Make bge_stop_fw() static. While I'm here use ANSI function definitions.
2010-07-13 19:33:46 +00:00
bge_sig_pre_reset(struct bge_softc *sc, int type)
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
{
Make bge_stop_fw() static. While I'm here use ANSI function definitions.
2010-07-13 19:33:46 +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
/*
* 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
Make bge_stop_fw() static. While I'm here use ANSI function definitions.
2010-07-13 19:33:46 +00:00
bge_sig_post_reset(struct bge_softc *sc, int type)
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
{
Make bge_stop_fw() static. While I'm here use ANSI function definitions.
2010-07-13 19:33:46 +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
if (sc->bge_asf_mode & ASF_NEW_HANDSHAKE) {
switch (type) {
case BGE_RESET_START:
Remove extra white space.
2009-11-22 18:30:19 +00:00
bge_writemem_ind(sc, BGE_SDI_STATUS, 0x80000001);
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
/* START DONE */
break;
case BGE_RESET_STOP:
Remove extra white space.
2009-11-22 18:30:19 +00:00
bge_writemem_ind(sc, BGE_SDI_STATUS, 0x80000002);
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
break;
}
}
}
static void
Make bge_stop_fw() static. While I'm here use ANSI function definitions.
2010-07-13 19:33:46 +00:00
bge_sig_legacy(struct bge_softc *sc, int type)
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
{
Make bge_stop_fw() static. While I'm here use ANSI function definitions.
2010-07-13 19:33:46 +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
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;
}
}
}
Make bge_stop_fw() static. While I'm here use ANSI function definitions.
2010-07-13 19:33:46 +00:00
static void
bge_stop_fw(struct bge_softc *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
{
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,
Use correct bits to stop firmware when ASF mode is enabled.
2007-03-08 00:49:26 +00:00
CSR_READ_4(sc, BGE_CPU_EVENT) | (1 << 14));
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
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
/*
- Ensure that INTx isn't disabled, as these chips apparently have a quirk requiring it to be enabled even when using MSI. This makes the latter work again after r189285. - Remove a comment which no longer applies since r190194.
2009-03-23 14:36:50 +00:00
* Do endian, PCI and DMA initialization.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_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;
Enable hardware fixes for BCM5704 B0 as recommended by data sheet.
2010-03-10 20:22:57 +00:00
uint16_t val;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +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
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
/* 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);
Enable hardware fixes for BCM5704 B0 as recommended by data sheet.
2010-03-10 20:22:57 +00:00
if (sc->bge_chiprev == BGE_CHIPREV_5704_BX) {
/*
Fix typo in r204975. Pointed out by: marius
2010-03-10 20:55:55 +00:00
* Fix data corruption caused by non-qword write with WB.
Enable hardware fixes for BCM5704 B0 as recommended by data sheet.
2010-03-10 20:22:57 +00:00
* Fix master abort in PCI mode.
* Fix PCI latency timer.
*/
val = pci_read_config(sc->bge_dev, BGE_PCI_MSI_DATA + 2, 2);
val |= (1 << 10) | (1 << 12) | (1 << 13);
pci_write_config(sc->bge_dev, BGE_PCI_MSI_DATA + 2, val, 2);
}
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
/*
* Set up the PCI DMA control register.
*/
dma_rw_ctl = BGE_PCIDMARWCTL_RD_CMD_SHIFT(6) |
BGE_PCIDMARWCTL_WR_CMD_SHIFT(7);
Pack several boolean fields into single bge_flags field.
2006-08-23 11:32:54 +00:00
if (sc->bge_flags & BGE_FLAG_PCIE) {
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
/* Read watermark not used, 128 bytes for write. */
dma_rw_ctl |= BGE_PCIDMARWCTL_WR_WAT_SHIFT(3);
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) {
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)) {
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
/* 256 bytes for read and write. */
dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(2) |
BGE_PCIDMARWCTL_WR_WAT_SHIFT(2);
dma_rw_ctl |= (sc->bge_asicrev == BGE_ASICREV_BCM5780) ?
BGE_PCIDMARWCTL_ONEDMA_ATONCE_GLOBAL :
BGE_PCIDMARWCTL_ONEDMA_ATONCE_LOCAL;
Set maximum read byte count to 2048 for PCI-X BCM5703/5704 devices. Also disable relaxed ordering as recommended by data sheet for PCI-X devices. For PCI-X BCM5704, set maximum outstanding split transactions to 0 as indicated by data sheet. For BCM5703 in PCI-X mode, DMA read watermark should be less than or equal to maximum read byte count configuration. Enforce this limitation in DMA read watermark configuration.
2010-03-10 20:54:08 +00:00
} else if (sc->bge_asicrev == BGE_ASICREV_BCM5703) {
/*
* In the BCM5703, the DMA read watermark should
* be set to less than or equal to the maximum
* memory read byte count of the PCI-X command
* register.
*/
dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(4) |
BGE_PCIDMARWCTL_WR_WAT_SHIFT(3);
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
} else if (sc->bge_asicrev == BGE_ASICREV_BCM5704) {
/* 1536 bytes for read, 384 bytes for write. */
dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(7) |
BGE_PCIDMARWCTL_WR_WAT_SHIFT(3);
} else {
/* 384 bytes for read and write. */
dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(3) |
BGE_PCIDMARWCTL_WR_WAT_SHIFT(3) |
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
0x0F;
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +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 ||
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
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
/* Set ONE_DMA_AT_ONCE for hardware workaround. */
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
tmp = CSR_READ_4(sc, BGE_PCI_CLKCTL) & 0x1F;
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
if (tmp == 6 || tmp == 7)
dma_rw_ctl |=
BGE_PCIDMARWCTL_ONEDMA_ATONCE_GLOBAL;
/* Set PCI-X DMA write workaround. */
dma_rw_ctl |= BGE_PCIDMARWCTL_ASRT_ALL_BE;
- 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
}
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
} else {
/* Conventional PCI bus: 256 bytes for read and write. */
dma_rw_ctl |= BGE_PCIDMARWCTL_RD_WAT_SHIFT(7) |
BGE_PCIDMARWCTL_WR_WAT_SHIFT(7);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
if (sc->bge_asicrev != BGE_ASICREV_BCM5705 &&
sc->bge_asicrev != BGE_ASICREV_BCM5750)
dma_rw_ctl |= 0x0F;
}
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 ||
sc->bge_asicrev == BGE_ASICREV_BCM5701)
dma_rw_ctl |= BGE_PCIDMARWCTL_USE_MRM |
BGE_PCIDMARWCTL_ASRT_ALL_BE;
- 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 ||
Rearrange DMA read/write control register settings based on document snippet provided by davidch via glebius. PR: kern/96806
2007-05-22 19:22:58 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5704)
- Change the short hand representation of the various ASIC revisions - Implement the ONEDMA_AT_ONCE workaround as described in the 5703/5704 eratta documents. Obtained from: NetBSD & Broadcom documentation
2003-05-03 22:58:45 +00:00
dma_rw_ctl &= ~BGE_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.
*/
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
CSR_WRITE_4(sc, BGE_MODE_CTL, BGE_DMA_SWAP_OPTIONS |
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
- Limit BCM5701 B5 to 32-bit mode as a workaround for a bug which causes data corruption in combination with certain bridges. Information about this problem was kindly provided by davidch. [1] - As BGE_FLAG_PCIX is meant to indicate that the controller is in PCI-X mode, revert to the pre __FreeBSD_version 602101 method of reading the bus mode register rather than checking the mere existence of a PCI-X capability, which is also there when the NIC f.e. is put into a 32-bit slot causing it not to be in PCI-X mode. Setting BGE_FLAG_PCIX inappropriately could cause the NIC to be tuned incorrectly. PR: 128833 [1] Reviewed by: jhb MFC after: 3 days
2008-12-09 21:34:22 +00:00
/*
* BCM5701 B5 have a bug causing data corruption when using
* 64-bit DMA reads, which can be terminated early and then
* completed later as 32-bit accesses, in combination with
* certain bridges.
*/
if (sc->bge_asicrev == BGE_ASICREV_BCM5701 &&
sc->bge_chipid == BGE_CHIPID_BCM5701_B5)
BGE_SETBIT(sc, BGE_MODE_CTL, BGE_MODECTL_FORCE_PCI32);
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
- Ensure that INTx isn't disabled, as these chips apparently have a quirk requiring it to be enabled even when using MSI. This makes the latter work again after r189285. - Remove a comment which no longer applies since r190194.
2009-03-23 14:36:50 +00:00
* properly by these devices. Also ensure that INTx isn't disabled,
* as these chips need it even when using MSI.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
*/
- Ensure that INTx isn't disabled, as these chips apparently have a quirk requiring it to be enabled even when using MSI. This makes the latter work again after r189285. - Remove a comment which no longer applies since r190194.
2009-03-23 14:36:50 +00:00
PCI_CLRBIT(sc->bge_dev, BGE_PCI_CMD,
PCIM_CMD_INTxDIS | 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
/* Set the timer prescaler (always 66Mhz) */
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
CSR_WRITE_4(sc, BGE_MISC_CFG, BGE_32BITTIME_66MHZ);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
/* XXX: The Linux tg3 driver does this at the start of brgphy_reset. */
if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
DELAY(40); /* XXX */
/* Put PHY into ready state */
BGE_CLRBIT(sc, BGE_MISC_CFG, BGE_MISCCFG_EPHY_IDDQ);
CSR_READ_4(sc, BGE_MISC_CFG); /* Flush */
DELAY(40);
}
- 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 support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +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, 0x50);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x20);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0x60);
} else if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x0);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x04);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0x10);
} else {
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_READDMA_LOWAT, 0x0);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_MACRX_LOWAT, 0x10);
CSR_WRITE_4(sc, BGE_BMAN_MBUFPOOL_HIWAT, 0x60);
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
/* 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,
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
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++) {
Fix time out check in EEPROM read and move delays to give some settle time. Submitted by: mjacob
2007-05-22 18:51:05 +00:00
DELAY(10);
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
if (CSR_READ_4(sc, BGE_BMAN_MODE) & BGE_BMANMODE_ENABLE)
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
break;
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
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 */
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
CSR_WRITE_4(sc, BGE_FTQ_RESET, 0xFFFFFFFF);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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, 0);
/* Wait until queue initialization is complete */
for (i = 0; i < BGE_TIMEOUT; i++) {
Fix time out check in EEPROM read and move delays to give some settle time. Submitted by: mjacob
2007-05-22 18:51:05 +00:00
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
if (CSR_READ_4(sc, BGE_FTQ_RESET) == 0)
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, "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;
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +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,
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
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 */
Partially revert r200228. For mini RCB case, bge(4) still have to disable mini ring withtout regard to mini ring support. Reported by: marcel Tested by: marcel
2009-12-08 03:24:29 +00:00
rcb = &sc->bge_ldata.bge_info.bge_mini_rx_rcb;
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 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
/*
* Set the BD ring replentish thresholds. The recommended
* values are 1/8th the number of descriptors allocated to
* each ring.
Fix recent style problems and expand a comment. Submitted by: Bruce Evans
2006-12-13 21:13:09 +00:00
* XXX The 5754 requires a lower threshold, so it might be a
* requirement of all 575x family chips. The Linux driver sets
* the lower threshold for all 5705 family chips as well, but there
* are reports that it might not need to be so strict.
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
*
* XXX Linux does some extra fiddling here for the 5906 parts as
* well.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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);
Don't access jumbo frame related registers if controller lacks the feature. These registers are reserved on controllers that have no support for jumbo frame. Only BCM5700 has mini ring so do not poke mini ring related registers if controller is not BCM5700. Reviewed by: marius
2009-12-07 19:26:54 +00:00
if (BGE_IS_JUMBO_CAPABLE(sc))
CSR_WRITE_4(sc, BGE_RBDI_JUMBO_REPL_THRESH,
BGE_JUMBO_RX_RING_CNT/8);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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 support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(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 */
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_RX_STD_PROD_LO, 0);
Don't access jumbo frame related registers if controller lacks the feature. These registers are reserved on controllers that have no support for jumbo frame. Only BCM5700 has mini ring so do not poke mini ring related registers if controller is not BCM5700. Reviewed by: marius
2009-12-07 19:26:54 +00:00
if (BGE_IS_JUMBO_CAPABLE(sc))
bge_writembx(sc, BGE_MBX_RX_JUMBO_PROD_LO, 0);
if (sc->bge_asicrev == BGE_ASICREV_BCM5700)
bge_writembx(sc, BGE_MBX_RX_MINI_PROD_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
/*
* 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,
Remove extra white space.
2009-11-22 18:30:19 +00:00
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. */
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
CSR_WRITE_4(sc, BGE_RXLP_STATS_ENABLE_MASK, 0x007FFFFF);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_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++) {
Fix time out check in EEPROM read and move delays to give some settle time. Submitted by: mjacob
2007-05-22 18:51:05 +00:00
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
if (!(CSR_READ_4(sc, BGE_HCC_MODE) & BGE_HCCMODE_ENABLE))
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,
"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);
}
After rev.1.169, the "interrupt" coalescing parameters are not used in bge_intr(). Some of them are used in bge_poll(). Simplify by only initializing these for polling mode and not toggling them when switching modes. This also fixes missing synchronization with the coalescing engine in the toggling.
2006-12-26 18:33:55 +00:00
CSR_WRITE_4(sc, BGE_HCC_RX_MAX_COAL_BDS_INT, 1);
CSR_WRITE_4(sc, BGE_HCC_TX_MAX_COAL_BDS_INT, 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
/* 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));
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
Reduce status block size DMAed by controller. bge(4) uses single Tx/Rx/Rx return ring such that large part of status block was not used at all. All bge(4) controllers except BCM5700 AX/BX has a feature to control the size of status block. So use minimum status block size allowed in controller. This reduces number of DMAed status block size to 32 bytes from 80 bytes.
2009-11-22 21:45:55 +00:00
/* Set up status block size. */
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
Zero entire status block and add missing bus_dmamap_sync(9).
2010-07-06 18:17:31 +00:00
sc->bge_chipid != BGE_CHIPID_BCM5700_C0) {
Reduce status block size DMAed by controller. bge(4) uses single Tx/Rx/Rx return ring such that large part of status block was not used at all. All bge(4) controllers except BCM5700 AX/BX has a feature to control the size of status block. So use minimum status block size allowed in controller. This reduces number of DMAed status block size to 32 bytes from 80 bytes.
2009-11-22 21:45:55 +00:00
val = BGE_STATBLKSZ_FULL;
Zero entire status block and add missing bus_dmamap_sync(9).
2010-07-06 18:17:31 +00:00
bzero(sc->bge_ldata.bge_status_block, BGE_STATUS_BLK_SZ);
} else {
Reduce status block size DMAed by controller. bge(4) uses single Tx/Rx/Rx return ring such that large part of status block was not used at all. All bge(4) controllers except BCM5700 AX/BX has a feature to control the size of status block. So use minimum status block size allowed in controller. This reduces number of DMAed status block size to 32 bytes from 80 bytes.
2009-11-22 21:45:55 +00:00
val = BGE_STATBLKSZ_32BYTE;
Zero entire status block and add missing bus_dmamap_sync(9).
2010-07-06 18:17:31 +00:00
bzero(sc->bge_ldata.bge_status_block, 32);
}
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
Reduce status block size DMAed by controller. bge(4) uses single Tx/Rx/Rx return ring such that large part of status block was not used at all. All bge(4) controllers except BCM5700 AX/BX has a feature to control the size of status block. So use minimum status block size allowed in controller. This reduces number of DMAed status block size to 32 bytes from 80 bytes.
2009-11-22 21:45:55 +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 host coalescing state machine */
Reduce status block size DMAed by controller. bge(4) uses single Tx/Rx/Rx return ring such that large part of status block was not used at all. All bge(4) controllers except BCM5700 AX/BX has a feature to control the size of status block. So use minimum status block size allowed in controller. This reduces number of DMAed status block size to 32 bytes from 80 bytes.
2009-11-22 21:45:55 +00:00
CSR_WRITE_4(sc, BGE_HCC_MODE, val | BGE_HCCMODE_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 RX BD completion state machine and enable attentions */
CSR_WRITE_4(sc, BGE_RBDC_MODE,
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
BGE_RBDCMODE_ENABLE | BGE_RBDCMODE_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
/* 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
For controllers that has dual mode PHY(copper or fiber) interfaces over GMII, make sure to enable GMII. With this change brgphy(4) is used to handle the dual mode PHY. Since we still don't have a sane way to pass PHY specific information to mii(4) layer special handling is needed in brgphy(4) to determine which mode of PHY was configured in parent interface. This change make BCM5715S work. Tested by: olli Obtained from: OpenBSD MFC after: 1 week
2010-01-14 19:08:43 +00:00
val = 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;
if (sc->bge_flags & BGE_FLAG_TBI)
val |= BGE_PORTMODE_TBI;
else if (sc->bge_flags & BGE_FLAG_MII_SERDES)
val |= BGE_PORTMODE_GMII;
else
val |= 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
/* Turn on DMA, clear stats */
For controllers that has dual mode PHY(copper or fiber) interfaces over GMII, make sure to enable GMII. With this change brgphy(4) is used to handle the dual mode PHY. Since we still don't have a sane way to pass PHY specific information to mii(4) layer special handling is needed in brgphy(4) to determine which mode of PHY was configured in parent interface. This change make BCM5715S work. Tested by: olli Obtained from: OpenBSD MFC after: 1 week
2010-01-14 19:08:43 +00:00
CSR_WRITE_4(sc, BGE_MAC_MODE, 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
/* 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 */
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
/* 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 style(9) and consistency.
2007-03-06 20:14:48 +00:00
val = BGE_WDMAMODE_ENABLE | BGE_WDMAMODE_ALL_ATTNS;
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
/* Enable host coalescing bug fix. */
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
if (BGE_IS_5755_PLUS(sc))
- Give a name to the host coalescing bug fix WDMA mode register bit instead of using hardcoded value in the code. Obtained from: OpenBSD
2009-10-07 14:29:48 +00:00
val |= BGE_WDMAMODE_STATUS_TAG_FIX;
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
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
- In bge_ifmedia_upd_locked() take advantrage of LIST_FOREACH(). - If boot verbose, print asicrev, chiprev and bus type on attach. - For PCI Express devices: 1) Adjust max read request size to 4Kbytes 2) Turn on FIFO_LONG_BURST in RDMA during bge_blockinit() Though 1) does not seem to have much to do with the poor TX performance observed on PCI Express bge(4), 2) does fix the problem. [1] - Nuke the RX CPU self-diag, which prevents working cards from working (Linux tg3 does not have this diag neither does OpenBSD's bge(4)). The increasing of the firmware handshaking timeout to 20000 retries done as part of the original commit isn't merged as way already have a way higher BGE_TIMEOUT of 100000. PR: 119361 [1] Obtained from: tg3 via DragonflyBSD [1], DragonflyBSD
2009-03-21 00:23:07 +00:00
DELAY(40);
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 */
- In bge_ifmedia_upd_locked() take advantrage of LIST_FOREACH(). - If boot verbose, print asicrev, chiprev and bus type on attach. - For PCI Express devices: 1) Adjust max read request size to 4Kbytes 2) Turn on FIFO_LONG_BURST in RDMA during bge_blockinit() Though 1) does not seem to have much to do with the poor TX performance observed on PCI Express bge(4), 2) does fix the problem. [1] - Nuke the RX CPU self-diag, which prevents working cards from working (Linux tg3 does not have this diag neither does OpenBSD's bge(4)). The increasing of the firmware handshaking timeout to 20000 retries done as part of the original commit isn't merged as way already have a way higher BGE_TIMEOUT of 100000. PR: 119361 [1] Obtained from: tg3 via DragonflyBSD [1], DragonflyBSD
2009-03-21 00:23:07 +00:00
val = BGE_RDMAMODE_ENABLE | BGE_RDMAMODE_ALL_ATTNS;
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
sc->bge_asicrev == BGE_ASICREV_BCM57780)
val |= BGE_RDMAMODE_BD_SBD_CRPT_ATTN |
BGE_RDMAMODE_MBUF_RBD_CRPT_ATTN |
BGE_RDMAMODE_MBUF_SBD_CRPT_ATTN;
- In bge_ifmedia_upd_locked() take advantrage of LIST_FOREACH(). - If boot verbose, print asicrev, chiprev and bus type on attach. - For PCI Express devices: 1) Adjust max read request size to 4Kbytes 2) Turn on FIFO_LONG_BURST in RDMA during bge_blockinit() Though 1) does not seem to have much to do with the poor TX performance observed on PCI Express bge(4), 2) does fix the problem. [1] - Nuke the RX CPU self-diag, which prevents working cards from working (Linux tg3 does not have this diag neither does OpenBSD's bge(4)). The increasing of the firmware handshaking timeout to 20000 retries done as part of the original commit isn't merged as way already have a way higher BGE_TIMEOUT of 100000. PR: 119361 [1] Obtained from: tg3 via DragonflyBSD [1], DragonflyBSD
2009-03-21 00:23:07 +00:00
if (sc->bge_flags & BGE_FLAG_PCIE)
val |= BGE_RDMAMODE_FIFO_LONG_BURST;
It seems read DMA mode register requires both IPv4 TSO and IPv6 TSO configuration to get IPv4 TSO work on BCM57780. While I'm here apply the same fix to BCM5785 which shares similar hardware feature of BCM57780. This change makes TSO work on BCM57780. Tested by: Tong Liu <nemoliu <> gmail dot com>
2010-07-06 02:07:59 +00:00
if (sc->bge_flags & BGE_FLAG_TSO) {
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
val |= BGE_RDMAMODE_TSO4_ENABLE;
It seems read DMA mode register requires both IPv4 TSO and IPv6 TSO configuration to get IPv4 TSO work on BCM57780. While I'm here apply the same fix to BCM5785 which shares similar hardware feature of BCM57780. This change makes TSO work on BCM57780. Tested by: Tong Liu <nemoliu <> gmail dot com>
2010-07-06 02:07:59 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5785 ||
sc->bge_asicrev == BGE_ASICREV_BCM57780)
val |= BGE_RDMAMODE_TSO6_ENABLE;
}
- In bge_ifmedia_upd_locked() take advantrage of LIST_FOREACH(). - If boot verbose, print asicrev, chiprev and bus type on attach. - For PCI Express devices: 1) Adjust max read request size to 4Kbytes 2) Turn on FIFO_LONG_BURST in RDMA during bge_blockinit() Though 1) does not seem to have much to do with the poor TX performance observed on PCI Express bge(4), 2) does fix the problem. [1] - Nuke the RX CPU self-diag, which prevents working cards from working (Linux tg3 does not have this diag neither does OpenBSD's bge(4)). The increasing of the firmware handshaking timeout to 20000 retries done as part of the original commit isn't merged as way already have a way higher BGE_TIMEOUT of 100000. PR: 119361 [1] Obtained from: tg3 via DragonflyBSD [1], DragonflyBSD
2009-03-21 00:23:07 +00:00
CSR_WRITE_4(sc, BGE_RDMA_MODE, val);
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
/* 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 */
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
val = BGE_SDCMODE_ENABLE;
if (sc->bge_asicrev == BGE_ASICREV_BCM5761)
val |= BGE_SDCMODE_CDELAY;
CSR_WRITE_4(sc, BGE_SDC_MODE, 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
/* Turn on send data initiator state machine */
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
if (sc->bge_flags & BGE_FLAG_TSO)
CSR_WRITE_4(sc, BGE_SDI_MODE, BGE_SDIMODE_ENABLE | 0x08);
else
CSR_WRITE_4(sc, BGE_SDI_MODE, BGE_SDIMODE_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 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);
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
CSR_WRITE_4(sc, BGE_SDI_STATS_ENABLE_MASK, 0x007FFFFF);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_SDI_STATS_CTL,
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
BGE_SDISTATSCTL_ENABLE | BGE_SDISTATSCTL_FASTER);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* ack/clear link change events */
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED |
BGE_MACSTAT_CFG_CHANGED | BGE_MACSTAT_MI_COMPLETE |
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_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 {
Fix style(9) and consistency.
2007-03-06 20:14:48 +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.
*/
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED |
BGE_MACSTAT_CFG_CHANGED | BGE_MACSTAT_MI_COMPLETE |
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
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);
Remove extra white space.
2009-11-22 18:30:19 +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
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
- Remove stale VPD support and its comment and get device name from VPD API. - Do not repeatedly read vendor/device IDs while probing. - Remove redundant bzero(3) for softc. device_get_softc(9) does it for free[1]. Reviewed by: glebius Suggested by: glebius[1]
2006-12-18 16:40:04 +00:00
* try to get the device name string from the controller itself instead
* of the compiled-in string. It guarantees we'll always announce the
* right product name. We fall back to the compiled-in string when
* VPD is unavailable or corrupt.
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
{
- Don't read the identifier string from the VPD if there's no chip containing an Ethernet address fitted as this is yet another thing that fails in that case in order to avoid the one second delay until pci_read_vpd_reg() times out. - Const'ify the bge_devs array.
2008-09-22 20:55:19 +00:00
const struct bge_type *t = bge_devs;
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_softc *sc = device_get_softc(dev);
- Remove stale VPD support and its comment and get device name from VPD API. - Do not repeatedly read vendor/device IDs while probing. - Remove redundant bzero(3) for softc. device_get_softc(9) does it for free[1]. Reviewed by: glebius Suggested by: glebius[1]
2006-12-18 16:40:04 +00:00
uint16_t vid, did;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_dev = dev;
- Remove stale VPD support and its comment and get device name from VPD API. - Do not repeatedly read vendor/device IDs while probing. - Remove redundant bzero(3) for softc. device_get_softc(9) does it for free[1]. Reviewed by: glebius Suggested by: glebius[1]
2006-12-18 16:40:04 +00:00
vid = pci_get_vendor(dev);
did = pci_get_device(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) {
- Remove stale VPD support and its comment and get device name from VPD API. - Do not repeatedly read vendor/device IDs while probing. - Remove redundant bzero(3) for softc. device_get_softc(9) does it for free[1]. Reviewed by: glebius Suggested by: glebius[1]
2006-12-18 16:40:04 +00:00
if ((vid == t->bge_vid) && (did == t->bge_did)) {
char model[64], buf[96];
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 *br;
const struct bge_vendor *v;
uint32_t id;
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
id = pci_read_config(dev, BGE_PCI_MISC_CTL, 4) >>
BGE_PCIMISCCTL_ASICREV_SHIFT;
if (BGE_ASICREV(id) == BGE_ASICREV_USE_PRODID_REG)
id = pci_read_config(dev,
BGE_PCI_PRODID_ASICREV, 4);
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
br = bge_lookup_rev(id);
- Remove stale VPD support and its comment and get device name from VPD API. - Do not repeatedly read vendor/device IDs while probing. - Remove redundant bzero(3) for softc. device_get_softc(9) does it for free[1]. Reviewed by: glebius Suggested by: glebius[1]
2006-12-18 16:40:04 +00:00
v = bge_lookup_vendor(vid);
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
{
#if __FreeBSD_version > 700024
const char *pname;
- Don't read the identifier string from the VPD if there's no chip containing an Ethernet address fitted as this is yet another thing that fails in that case in order to avoid the one second delay until pci_read_vpd_reg() times out. - Const'ify the bge_devs array.
2008-09-22 20:55:19 +00:00
if (bge_has_eaddr(sc) &&
pci_get_vpd_ident(dev, &pname) == 0)
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
snprintf(model, 64, "%s", pname);
else
#endif
snprintf(model, 64, "%s %s",
v->v_name,
br != NULL ? br->br_name :
"NetXtreme Ethernet Controller");
}
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
snprintf(buf, 96, "%s, %sASIC rev. %#08x", model,
br != NULL ? "" : "unknown ", id);
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
device_set_desc_copy(dev, buf);
- 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])
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_destroy(sc->bge_cdata.bge_rx_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
sc->bge_cdata.bge_rx_std_dmamap[i]);
}
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
if (sc->bge_cdata.bge_rx_std_sparemap)
bus_dmamap_destroy(sc->bge_cdata.bge_rx_mtag,
sc->bge_cdata.bge_rx_std_sparemap);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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
/* 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]);
}
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
if (sc->bge_cdata.bge_rx_jumbo_sparemap)
bus_dmamap_destroy(sc->bge_cdata.bge_mtag_jumbo,
sc->bge_cdata.bge_rx_jumbo_sparemap);
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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
/* 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])
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_destroy(sc->bge_cdata.bge_tx_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
sc->bge_cdata.bge_tx_dmamap[i]);
}
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
if (sc->bge_cdata.bge_rx_mtag)
bus_dma_tag_destroy(sc->bge_cdata.bge_rx_mtag);
if (sc->bge_cdata.bge_tx_mtag)
bus_dma_tag_destroy(sc->bge_cdata.bge_tx_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
- 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);
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
if (sc->bge_cdata.bge_buffer_tag)
bus_dma_tag_destroy(sc->bge_cdata.bge_buffer_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
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
bge_dma_ring_alloc(struct bge_softc *sc, bus_size_t alignment,
bus_size_t maxsize, bus_dma_tag_t *tag, uint8_t **ring, bus_dmamap_t *map,
bus_addr_t *paddr, const char *msg)
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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;
Fix two long standing bugs on bge(4). Most pre BCM5755 controllers have a DMA bug when buffer address crosses a multiple of the 4GB boundary(e.g. 4GB, 8GB, 12GB etc). Limit DMA address to be within 4GB address for these controllers. The second DMA bug limits DMA address to be within 40bit address space. This bug applies to BCM5714 and BCM5715 and 5708(bce(4) controller). This is not actually a MAC controller bug but an issue with the embedded PCIe to PCI-X bridge in the device. So for BCM5714/BCM5715 controllers also limit the DMA address to be within 40bit address space. Special thanks to davidch@ who gave me detailed errata information. I think this change will fix long standing bge(4) instability issues on systems with more than 4GB memory. Reviewed by: davidch
2009-11-22 20:50:27 +00:00
bus_addr_t lowaddr;
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
bus_size_t ring_end;
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
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
lowaddr = BUS_SPACE_MAXADDR;
again:
error = bus_dma_tag_create(sc->bge_cdata.bge_parent_tag,
alignment, 0, lowaddr, BUS_SPACE_MAXADDR, NULL,
NULL, maxsize, 1, maxsize, 0, NULL, NULL, tag);
if (error != 0) {
device_printf(sc->bge_dev,
"could not create %s dma tag\n", msg);
return (ENOMEM);
}
/* Allocate DMA'able memory for ring. */
error = bus_dmamem_alloc(*tag, (void **)ring,
BUS_DMA_NOWAIT | BUS_DMA_ZERO | BUS_DMA_COHERENT, map);
if (error != 0) {
device_printf(sc->bge_dev,
"could not allocate DMA'able memory for %s\n", msg);
return (ENOMEM);
}
/* Load the address of the ring. */
ctx.bge_busaddr = 0;
error = bus_dmamap_load(*tag, *map, *ring, maxsize, bge_dma_map_addr,
&ctx, BUS_DMA_NOWAIT);
if (error != 0) {
device_printf(sc->bge_dev,
"could not load DMA'able memory for %s\n", msg);
return (ENOMEM);
}
*paddr = ctx.bge_busaddr;
ring_end = *paddr + maxsize;
if ((sc->bge_flags & BGE_FLAG_4G_BNDRY_BUG) != 0 &&
BGE_ADDR_HI(*paddr) != BGE_ADDR_HI(ring_end)) {
/*
* 4GB boundary crossed. Limit maximum allowable DMA
* address space to 32bit and try again.
*/
bus_dmamap_unload(*tag, *map);
bus_dmamem_free(*tag, *ring, *map);
bus_dma_tag_destroy(*tag);
if (bootverbose)
device_printf(sc->bge_dev, "4GB boundary crossed, "
"limit DMA address space to 32bit for %s\n", msg);
*ring = NULL;
*tag = NULL;
*map = NULL;
lowaddr = BUS_SPACE_MAXADDR_32BIT;
goto again;
}
return (0);
}
static int
bge_dma_alloc(struct bge_softc *sc)
{
bus_addr_t lowaddr;
bus_size_t boundary, sbsz, txsegsz, txmaxsegsz;
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
Fix two long standing bugs on bge(4). Most pre BCM5755 controllers have a DMA bug when buffer address crosses a multiple of the 4GB boundary(e.g. 4GB, 8GB, 12GB etc). Limit DMA address to be within 4GB address for these controllers. The second DMA bug limits DMA address to be within 40bit address space. This bug applies to BCM5714 and BCM5715 and 5708(bce(4) controller). This is not actually a MAC controller bug but an issue with the embedded PCIe to PCI-X bridge in the device. So for BCM5714/BCM5715 controllers also limit the DMA address to be within 40bit address space. Special thanks to davidch@ who gave me detailed errata information. I think this change will fix long standing bge(4) instability issues on systems with more than 4GB memory. Reviewed by: davidch
2009-11-22 20:50:27 +00:00
lowaddr = BUS_SPACE_MAXADDR;
if ((sc->bge_flags & BGE_FLAG_40BIT_BUG) != 0)
lowaddr = BGE_DMA_MAXADDR;
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since I don't have access to a non-x86 FreeBSD box at the moment. Volunteers welcome.
2003-07-25 19:42:44 +00:00
/*
* Allocate the parent bus DMA tag appropriate for PCI.
*/
- Use more appropriate maxsize, nsegments and maxsegsize parameters when creating the parent bus DMA tag. While at it correct the style and a nearby comment. - Take advantage of m_collapse(9) for performance reasons. MFC after: 2 weeks
2008-04-26 10:54:17 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(sc->bge_dev),
Fix two long standing bugs on bge(4). Most pre BCM5755 controllers have a DMA bug when buffer address crosses a multiple of the 4GB boundary(e.g. 4GB, 8GB, 12GB etc). Limit DMA address to be within 4GB address for these controllers. The second DMA bug limits DMA address to be within 40bit address space. This bug applies to BCM5714 and BCM5715 and 5708(bce(4) controller). This is not actually a MAC controller bug but an issue with the embedded PCIe to PCI-X bridge in the device. So for BCM5714/BCM5715 controllers also limit the DMA address to be within 40bit address space. Special thanks to davidch@ who gave me detailed errata information. I think this change will fix long standing bge(4) instability issues on systems with more than 4GB memory. Reviewed by: davidch
2009-11-22 20:50:27 +00:00
1, 0, lowaddr, BUS_SPACE_MAXADDR, NULL,
- Use more appropriate maxsize, nsegments and maxsegsize parameters when creating the parent bus DMA tag. While at it correct the style and a nearby comment. - Take advantage of m_collapse(9) for performance reasons. MFC after: 2 weeks
2008-04-26 10:54:17 +00:00
NULL, BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT,
0, NULL, NULL, &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);
}
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
/* Create tag for standard RX ring. */
error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_STD_RX_RING_SZ,
&sc->bge_cdata.bge_rx_std_ring_tag,
(uint8_t **)&sc->bge_ldata.bge_rx_std_ring,
&sc->bge_cdata.bge_rx_std_ring_map,
&sc->bge_ldata.bge_rx_std_ring_paddr, "RX ring");
if (error)
return (error);
/* Create tag for RX return ring. */
error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_RX_RTN_RING_SZ(sc),
&sc->bge_cdata.bge_rx_return_ring_tag,
(uint8_t **)&sc->bge_ldata.bge_rx_return_ring,
&sc->bge_cdata.bge_rx_return_ring_map,
&sc->bge_ldata.bge_rx_return_ring_paddr, "RX return ring");
if (error)
return (error);
/* Create tag for TX ring. */
error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_TX_RING_SZ,
&sc->bge_cdata.bge_tx_ring_tag,
(uint8_t **)&sc->bge_ldata.bge_tx_ring,
&sc->bge_cdata.bge_tx_ring_map,
&sc->bge_ldata.bge_tx_ring_paddr, "TX ring");
if (error)
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
/*
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
* Create tag for status block.
* Because we only use single Tx/Rx/Rx return ring, use
* minimum status block size except BCM5700 AX/BX which
* seems to want to see full status block size regardless
* of configured number of 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
*/
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 &&
sc->bge_chipid != BGE_CHIPID_BCM5700_C0)
sbsz = BGE_STATUS_BLK_SZ;
else
sbsz = 32;
error = bge_dma_ring_alloc(sc, PAGE_SIZE, sbsz,
&sc->bge_cdata.bge_status_tag,
(uint8_t **)&sc->bge_ldata.bge_status_block,
&sc->bge_cdata.bge_status_map,
&sc->bge_ldata.bge_status_block_paddr, "status block");
if (error)
return (error);
/* Create tag for jumbo RX ring. */
if (BGE_IS_JUMBO_CAPABLE(sc)) {
error = bge_dma_ring_alloc(sc, PAGE_SIZE, BGE_JUMBO_RX_RING_SZ,
&sc->bge_cdata.bge_rx_jumbo_ring_tag,
(uint8_t **)&sc->bge_ldata.bge_rx_jumbo_ring,
&sc->bge_cdata.bge_rx_jumbo_ring_map,
&sc->bge_ldata.bge_rx_jumbo_ring_paddr, "jumbo RX ring");
if (error)
return (error);
}
/* Create parent tag for buffers. */
boundary = 0;
if ((sc->bge_flags & BGE_FLAG_4G_BNDRY_BUG) != 0)
Handle PAE case correctly. You cannot effectively specify a 4GB boundary in PAE case so use a 2GB boundary for PAE as suggested by jhb. Pointed out by: jhb Reviewed by: jhb
2010-08-31 18:48:09 +00:00
boundary = BGE_DMA_BNDRY;
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
error = bus_dma_tag_create(bus_get_dma_tag(sc->bge_dev),
1, boundary, lowaddr, BUS_SPACE_MAXADDR, NULL,
NULL, BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT,
0, NULL, NULL, &sc->bge_cdata.bge_buffer_tag);
if (error != 0) {
device_printf(sc->bge_dev,
"could not allocate buffer dma tag\n");
return (ENOMEM);
}
/* Create tag for Tx mbufs. */
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
if (sc->bge_flags & BGE_FLAG_TSO) {
txsegsz = BGE_TSOSEG_SZ;
txmaxsegsz = 65535 + sizeof(struct ether_vlan_header);
} else {
txsegsz = MCLBYTES;
txmaxsegsz = MCLBYTES * BGE_NSEG_NEW;
}
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
error = bus_dma_tag_create(sc->bge_cdata.bge_buffer_tag, 1,
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
txmaxsegsz, BGE_NSEG_NEW, txsegsz, 0, NULL, NULL,
&sc->bge_cdata.bge_tx_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) {
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
device_printf(sc->bge_dev, "could not allocate TX dma tag\n");
return (ENOMEM);
}
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
/* Create tag for Rx mbufs. */
error = bus_dma_tag_create(sc->bge_cdata.bge_buffer_tag, 1, 0,
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, MCLBYTES, 1,
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
MCLBYTES, 0, NULL, NULL, &sc->bge_cdata.bge_rx_mtag);
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
if (error) {
device_printf(sc->bge_dev, "could not allocate RX 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. */
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
error = bus_dmamap_create(sc->bge_cdata.bge_rx_mtag, 0,
&sc->bge_cdata.bge_rx_std_sparemap);
if (error) {
device_printf(sc->bge_dev,
"can't create spare DMA map for 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
for (i = 0; i < BGE_STD_RX_RING_CNT; i++) {
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
error = bus_dmamap_create(sc->bge_cdata.bge_rx_mtag, 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
&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++) {
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
error = bus_dmamap_create(sc->bge_cdata.bge_tx_mtag, 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
&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,
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
"can't create DMA map for TX\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
}
}
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
/* Create tags for jumbo RX buffers. */
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)) {
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
error = bus_dma_tag_create(sc->bge_cdata.bge_buffer_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 DMA maps for jumbo RX buffers. */
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
error = bus_dmamap_create(sc->bge_cdata.bge_mtag_jumbo,
0, &sc->bge_cdata.bge_rx_jumbo_sparemap);
if (error) {
device_printf(sc->bge_dev,
Fix typo introduced in r199011. Pointed out by: marius
2009-11-22 18:34:15 +00:00
"can't create spare DMA map for jumbo RX\n");
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +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
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
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
}
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
/*
* Return true if this device has more than one port.
*/
static int
bge_has_multiple_ports(struct bge_softc *sc)
{
device_t dev = sc->bge_dev;
Make the PCI code aware of PCI domains (aka PCI segments) so we can support machines having multiple independently numbered PCI domains and don't support reenumeration without ambiguity amongst the devices as seen by the OS and represented by PCI location strings. This includes introducing a function pci_find_dbsf(9) which works like pci_find_bsf(9) but additionally takes a domain number argument and limiting pci_find_bsf(9) to only search devices in domain 0 (the only domain in single-domain systems). Bge(4) and ofw_pcibus(4) are changed to use pci_find_dbsf(9) instead of pci_find_bsf(9) in order to no longer report false positives when searching for siblings and dupe devices in the same domain respectively. Along with this change the sole host-PCI bridge driver converted to actually make use of PCI domain support is uninorth(4), the others continue to use domain 0 only for now and need to be converted as appropriate later on. Note that this means that the format of the location strings as used by pciconf(8) has been changed and that consumers of <sys/pciio.h> potentially need to be recompiled. Suggested by: jhb Reviewed by: grehan, jhb, marcel Approved by: re (kensmith), jhb (PCI maintainer hat)
2007-09-30 11:05:18 +00:00
u_int b, d, f, fscan, s;
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
Make the PCI code aware of PCI domains (aka PCI segments) so we can support machines having multiple independently numbered PCI domains and don't support reenumeration without ambiguity amongst the devices as seen by the OS and represented by PCI location strings. This includes introducing a function pci_find_dbsf(9) which works like pci_find_bsf(9) but additionally takes a domain number argument and limiting pci_find_bsf(9) to only search devices in domain 0 (the only domain in single-domain systems). Bge(4) and ofw_pcibus(4) are changed to use pci_find_dbsf(9) instead of pci_find_bsf(9) in order to no longer report false positives when searching for siblings and dupe devices in the same domain respectively. Along with this change the sole host-PCI bridge driver converted to actually make use of PCI domain support is uninorth(4), the others continue to use domain 0 only for now and need to be converted as appropriate later on. Note that this means that the format of the location strings as used by pciconf(8) has been changed and that consumers of <sys/pciio.h> potentially need to be recompiled. Suggested by: jhb Reviewed by: grehan, jhb, marcel Approved by: re (kensmith), jhb (PCI maintainer hat)
2007-09-30 11:05:18 +00:00
d = pci_get_domain(dev);
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
b = pci_get_bus(dev);
s = pci_get_slot(dev);
f = pci_get_function(dev);
for (fscan = 0; fscan <= PCI_FUNCMAX; fscan++)
Make the PCI code aware of PCI domains (aka PCI segments) so we can support machines having multiple independently numbered PCI domains and don't support reenumeration without ambiguity amongst the devices as seen by the OS and represented by PCI location strings. This includes introducing a function pci_find_dbsf(9) which works like pci_find_bsf(9) but additionally takes a domain number argument and limiting pci_find_bsf(9) to only search devices in domain 0 (the only domain in single-domain systems). Bge(4) and ofw_pcibus(4) are changed to use pci_find_dbsf(9) instead of pci_find_bsf(9) in order to no longer report false positives when searching for siblings and dupe devices in the same domain respectively. Along with this change the sole host-PCI bridge driver converted to actually make use of PCI domain support is uninorth(4), the others continue to use domain 0 only for now and need to be converted as appropriate later on. Note that this means that the format of the location strings as used by pciconf(8) has been changed and that consumers of <sys/pciio.h> potentially need to be recompiled. Suggested by: jhb Reviewed by: grehan, jhb, marcel Approved by: re (kensmith), jhb (PCI maintainer hat)
2007-09-30 11:05:18 +00:00
if (fscan != f && pci_find_dbsf(d, b, s, fscan) != NULL)
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
return (1);
return (0);
}
/*
* Return true if MSI can be used with this device.
*/
static int
bge_can_use_msi(struct bge_softc *sc)
{
int can_use_msi = 0;
switch (sc->bge_asicrev) {
Do as the Linux tg3 driver does and enable MSI support also for the BCM5714 revision A0 when in a multi-port configuration and unconditionally for the remainder of the class of BCM575X and beyond chips. This was prodded by mav and is based on a suggestion and a patch submitted by jhb. Reviewed by: jhb MFC after: 2 months
2008-10-27 22:10:01 +00:00
case BGE_ASICREV_BCM5714_A0:
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
case BGE_ASICREV_BCM5714:
/*
Do as the Linux tg3 driver does and enable MSI support also for the BCM5714 revision A0 when in a multi-port configuration and unconditionally for the remainder of the class of BCM575X and beyond chips. This was prodded by mav and is based on a suggestion and a patch submitted by jhb. Reviewed by: jhb MFC after: 2 months
2008-10-27 22:10:01 +00:00
* Apparently, MSI doesn't work when these chips are
* configured in single-port mode.
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
*/
if (bge_has_multiple_ports(sc))
can_use_msi = 1;
break;
case BGE_ASICREV_BCM5750:
if (sc->bge_chiprev != BGE_CHIPREV_5750_AX &&
sc->bge_chiprev != BGE_CHIPREV_5750_BX)
can_use_msi = 1;
break;
Do as the Linux tg3 driver does and enable MSI support also for the BCM5714 revision A0 when in a multi-port configuration and unconditionally for the remainder of the class of BCM575X and beyond chips. This was prodded by mav and is based on a suggestion and a patch submitted by jhb. Reviewed by: jhb MFC after: 2 months
2008-10-27 22:10:01 +00:00
default:
if (BGE_IS_575X_PLUS(sc))
can_use_msi = 1;
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
}
return (can_use_msi);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
Use a better approach to force the interrupt which should work for all cards/modes. In addition to the intr forcing added with rev. 1.205 adopt the other places to use the same logic. We need to exclude a few chips/revisions (5700, 5788) from using the enhanced version and fall back to the old way as that is the only method they support. Tested by: phk Suggested by: davidch, Broadcom (thanks a lot for the help!) MFC after: 16 days
2008-05-05 18:42:17 +00:00
uint32_t hwcfg = 0, misccfg;
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
u_char eaddr[ETHER_ADDR_LEN];
Due to newly added PCIe capabilities fallback code for finding the PCIe capability did not work right on recent controllers. Remove FreeBSD 6.x support code. Reviewed by: marius
2009-11-22 18:47:56 +00:00
int error, msicount, reg, rid, trys;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
TASK_INIT(&sc->bge_intr_task, 0, bge_intr_task, 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
/*
* Map control/status registers.
*/
pci_enable_busmaster(dev);
Prefer PCIR_BAR macro over BGE_PCI_BAR0.
2010-07-13 19:45:40 +00:00
rid = PCIR_BAR(0);
Convert callers to the new bus_alloc_resource_any(9) API. Submitted by: Mark Santcroos <marks@ripe.net> Reviewed by: imp, dfr, bde
2004-03-17 17:50:55 +00:00
sc->bge_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
Remove some remnant alpha hacks. Approved by: PCI-maintainers (imp, jhb)
2008-04-26 14:13:48 +00:00
RF_ACTIVE);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
}
- In bge_ifmedia_upd_locked() take advantrage of LIST_FOREACH(). - If boot verbose, print asicrev, chiprev and bus type on attach. - For PCI Express devices: 1) Adjust max read request size to 4Kbytes 2) Turn on FIFO_LONG_BURST in RDMA during bge_blockinit() Though 1) does not seem to have much to do with the poor TX performance observed on PCI Express bge(4), 2) does fix the problem. [1] - Nuke the RX CPU self-diag, which prevents working cards from working (Linux tg3 does not have this diag neither does OpenBSD's bge(4)). The increasing of the firmware handshaking timeout to 20000 retries done as part of the original commit isn't merged as way already have a way higher BGE_TIMEOUT of 100000. PR: 119361 [1] Obtained from: tg3 via DragonflyBSD [1], DragonflyBSD
2009-03-21 00:23:07 +00:00
/* Save various chip information. */
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_chipid =
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
pci_read_config(dev, BGE_PCI_MISC_CTL, 4) >>
BGE_PCIMISCCTL_ASICREV_SHIFT;
if (BGE_ASICREV(sc->bge_chipid) == BGE_ASICREV_USE_PRODID_REG)
sc->bge_chipid = pci_read_config(dev, BGE_PCI_PRODID_ASICREV,
4);
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
sc->bge_asicrev = BGE_ASICREV(sc->bge_chipid);
sc->bge_chiprev = BGE_CHIPREV(sc->bge_chipid);
Add a flag for Ethernet@WireSpeed capability and correct chip revisions. The idea was taken from OpenBSD and cross-referenced with Linux driver.
2008-01-18 22:09:50 +00:00
/*
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
* Don't enable Ethernet@WireSpeed for the 5700, 5906, or the
Add a flag for Ethernet@WireSpeed capability and correct chip revisions. The idea was taken from OpenBSD and cross-referenced with Linux driver.
2008-01-18 22:09:50 +00:00
* 5705 A0 and A1 chips.
*/
if (sc->bge_asicrev != BGE_ASICREV_BCM5700 &&
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
sc->bge_asicrev != BGE_ASICREV_BCM5906 &&
Add a flag for Ethernet@WireSpeed capability and correct chip revisions. The idea was taken from OpenBSD and cross-referenced with Linux driver.
2008-01-18 22:09:50 +00:00
sc->bge_chipid != BGE_CHIPID_BCM5705_A0 &&
sc->bge_chipid != BGE_CHIPID_BCM5705_A1)
sc->bge_flags |= BGE_FLAG_WIRESPEED;
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
if (bge_has_eaddr(sc))
sc->bge_flags |= BGE_FLAG_EADDR;
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
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) {
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
case BGE_ASICREV_BCM5755:
case BGE_ASICREV_BCM5761:
case BGE_ASICREV_BCM5784:
case BGE_ASICREV_BCM5785:
case BGE_ASICREV_BCM5787:
case BGE_ASICREV_BCM57780:
sc->bge_flags |= BGE_FLAG_5755_PLUS | BGE_FLAG_575X_PLUS |
BGE_FLAG_5705_PLUS;
break;
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
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 */;
Some comment/space changes (FALLTHRU -> FALLTHROUGH, space after while).
2009-05-14 22:36:56 +00:00
/* FALLTHROUGH */
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
case BGE_ASICREV_BCM5750:
case BGE_ASICREV_BCM5752:
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
case BGE_ASICREV_BCM5906:
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
sc->bge_flags |= BGE_FLAG_575X_PLUS;
Some comment/space changes (FALLTHRU -> FALLTHROUGH, space after while).
2009-05-14 22:36:56 +00:00
/* FALLTHROUGH */
Use bge_flags to save chipset family and remove dead code while I am here.
2006-12-04 19:50:49 +00:00
case BGE_ASICREV_BCM5705:
sc->bge_flags |= BGE_FLAG_5705_PLUS;
break;
}
- Fix BCM5754 support found in Dell PowerEdge SC440. - Move some PHY bug detections from brgphy.c to if_bge.c. - Do not penalize working PHYs. - Re-arrange bge_flags roughly by their categories. - Fix minor style(9) nits. PR: kern/107257 Obtained from: OpenBSD Tested by: Mike Hibler <mike at flux dot utah dot edu>
2007-01-15 21:43:43 +00:00
/* Set various bug flags. */
Add BCM5701 A0/B0 CRC bug workaround. Magic values taken from Linux driver.
2007-02-12 23:58:52 +00:00
if (sc->bge_chipid == BGE_CHIPID_BCM5701_A0 ||
sc->bge_chipid == BGE_CHIPID_BCM5701_B0)
sc->bge_flags |= BGE_FLAG_CRC_BUG;
- Fix BCM5754 support found in Dell PowerEdge SC440. - Move some PHY bug detections from brgphy.c to if_bge.c. - Do not penalize working PHYs. - Re-arrange bge_flags roughly by their categories. - Fix minor style(9) nits. PR: kern/107257 Obtained from: OpenBSD Tested by: Mike Hibler <mike at flux dot utah dot edu>
2007-01-15 21:43:43 +00:00
if (sc->bge_chiprev == BGE_CHIPREV_5703_AX ||
sc->bge_chiprev == BGE_CHIPREV_5704_AX)
sc->bge_flags |= BGE_FLAG_ADC_BUG;
if (sc->bge_chipid == BGE_CHIPID_BCM5704_A0)
sc->bge_flags |= BGE_FLAG_5704_A0_BUG;
Move device specific flag configuration to attach routine. The softc obtained in device probe wouldn't be the same one used in device attach. Drivers should not assume any values stored in softc structure in probe routine will be available for its attach routine.
2010-02-09 19:12:06 +00:00
if (pci_get_subvendor(dev) == DELL_VENDORID)
sc->bge_flags |= BGE_FLAG_NO_3LED;
if (pci_get_device(dev) == BCOM_DEVICEID_BCM5755M)
sc->bge_flags |= BGE_FLAG_ADJUST_TRIM;
Add PHY DSP code for BCM5755M. Obtained from: OpenBSD
2007-02-12 22:51:25 +00:00
if (BGE_IS_5705_PLUS(sc) &&
!(sc->bge_flags & BGE_FLAG_ADJUST_TRIM)) {
- Fix BCM5754 support found in Dell PowerEdge SC440. - Move some PHY bug detections from brgphy.c to if_bge.c. - Do not penalize working PHYs. - Re-arrange bge_flags roughly by their categories. - Fix minor style(9) nits. PR: kern/107257 Obtained from: OpenBSD Tested by: Mike Hibler <mike at flux dot utah dot edu>
2007-01-15 21:43:43 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5755 ||
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5761 ||
sc->bge_asicrev == BGE_ASICREV_BCM5784 ||
Don't enable the workaround for the jitter bug on the 5722. Obtained from: Linux tg3 driver
2008-03-11 15:05:54 +00:00
sc->bge_asicrev == BGE_ASICREV_BCM5787) {
o Add PCI ID for BCM 5756. o Don't enable BGE_FLAG_BER_BUG on both 5722 and 5756, and based on their PCI IDs rather than their chip IDs. Reported by: several PC-BSD users via kmoore Reviewed by: yongari, imp, jhb, davidch Sponsored by: iXsystems, Inc. MFC after: 2 weeks
2010-01-13 22:39:39 +00:00
if (pci_get_device(dev) != BCOM_DEVICEID_BCM5722 &&
pci_get_device(dev) != BCOM_DEVICEID_BCM5756)
Don't enable the workaround for the jitter bug on the 5722. Obtained from: Linux tg3 driver
2008-03-11 15:05:54 +00:00
sc->bge_flags |= BGE_FLAG_JITTER_BUG;
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
} else if (sc->bge_asicrev != BGE_ASICREV_BCM5906)
- Fix BCM5754 support found in Dell PowerEdge SC440. - Move some PHY bug detections from brgphy.c to if_bge.c. - Do not penalize working PHYs. - Re-arrange bge_flags roughly by their categories. - Fix minor style(9) nits. PR: kern/107257 Obtained from: OpenBSD Tested by: Mike Hibler <mike at flux dot utah dot edu>
2007-01-15 21:43:43 +00:00
sc->bge_flags |= BGE_FLAG_BER_BUG;
}
Fix two long standing bugs on bge(4). Most pre BCM5755 controllers have a DMA bug when buffer address crosses a multiple of the 4GB boundary(e.g. 4GB, 8GB, 12GB etc). Limit DMA address to be within 4GB address for these controllers. The second DMA bug limits DMA address to be within 40bit address space. This bug applies to BCM5714 and BCM5715 and 5708(bce(4) controller). This is not actually a MAC controller bug but an issue with the embedded PCIe to PCI-X bridge in the device. So for BCM5714/BCM5715 controllers also limit the DMA address to be within 40bit address space. Special thanks to davidch@ who gave me detailed errata information. I think this change will fix long standing bge(4) instability issues on systems with more than 4GB memory. Reviewed by: davidch
2009-11-22 20:50:27 +00:00
/*
* All controllers that are not 5755 or higher have 4GB
* boundary DMA bug.
* Whenever an address crosses a multiple of the 4GB boundary
* (including 4GB, 8Gb, 12Gb, etc.) and makes the transition
* from 0xX_FFFF_FFFF to 0x(X+1)_0000_0000 an internal DMA
* state machine will lockup and cause the device to hang.
*/
if (BGE_IS_5755_PLUS(sc) == 0)
sc->bge_flags |= BGE_FLAG_4G_BNDRY_BUG;
Use a better approach to force the interrupt which should work for all cards/modes. In addition to the intr forcing added with rev. 1.205 adopt the other places to use the same logic. We need to exclude a few chips/revisions (5700, 5788) from using the enhanced version and fall back to the old way as that is the only method they support. Tested by: phk Suggested by: davidch, Broadcom (thanks a lot for the help!) MFC after: 16 days
2008-05-05 18:42:17 +00:00
/*
* We could possibly check for BCOM_DEVICEID_BCM5788 in bge_probe()
* but I do not know the DEVICEID for the 5788M.
*/
misccfg = CSR_READ_4(sc, BGE_MISC_CFG) & BGE_MISCCFG_BOARD_ID;
if (misccfg == BGE_MISCCFG_BOARD_ID_5788 ||
misccfg == BGE_MISCCFG_BOARD_ID_5788M)
sc->bge_flags |= BGE_FLAG_5788;
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
/*
* Some controllers seem to require a special firmware to use
* TSO. But the firmware is not available to FreeBSD and Linux
* claims that the TSO performed by the firmware is slower than
* hardware based TSO. Moreover the firmware based TSO has one
* known bug which can't handle TSO if ethernet header + IP/TCP
* header is greater than 80 bytes. The workaround for the TSO
* bug exist but it seems it's too expensive than not using
* TSO at all. Some hardwares also have the TSO bug so limit
* the TSO to the controllers that are not affected TSO issues
* (e.g. 5755 or higher).
*/
Fix regression introduced in r198318. BCM5754/BCM5754M uses the same ASIC ID of BCM5758 such that r198318 incorecctly enabled TSO on BCM5754.BCM5754M controllers. BCM5754/BCM5754M needs a special firmware to enable TSO and bge(4) does not support firmware based TSO. Reported by: ed Tested by: ed
2010-01-03 21:49:24 +00:00
if (BGE_IS_5755_PLUS(sc)) {
/*
* BCM5754 and BCM5787 shares the same ASIC id so
* explicit device id check is required.
Disable TSO on BCM5755M controller until I understand better for the issue. I still have no idea why TSO does not work on this controller. davidch@ also confirmed there is no known TSO related issues for this controller.
2010-02-26 22:29:42 +00:00
* Due to unknown reason TSO does not work on BCM5755M.
Fix regression introduced in r198318. BCM5754/BCM5754M uses the same ASIC ID of BCM5758 such that r198318 incorecctly enabled TSO on BCM5754.BCM5754M controllers. BCM5754/BCM5754M needs a special firmware to enable TSO and bge(4) does not support firmware based TSO. Reported by: ed Tested by: ed
2010-01-03 21:49:24 +00:00
*/
if (pci_get_device(dev) != BCOM_DEVICEID_BCM5754 &&
Disable TSO on BCM5755M controller until I understand better for the issue. I still have no idea why TSO does not work on this controller. davidch@ also confirmed there is no known TSO related issues for this controller.
2010-02-26 22:29:42 +00:00
pci_get_device(dev) != BCOM_DEVICEID_BCM5754M &&
pci_get_device(dev) != BCOM_DEVICEID_BCM5755M)
Fix regression introduced in r198318. BCM5754/BCM5754M uses the same ASIC ID of BCM5758 such that r198318 incorecctly enabled TSO on BCM5754.BCM5754M controllers. BCM5754/BCM5754M needs a special firmware to enable TSO and bge(4) does not support firmware based TSO. Reported by: ed Tested by: ed
2010-01-03 21:49:24 +00:00
sc->bge_flags |= BGE_FLAG_TSO;
}
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +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
/*
* Check if this is a PCI-X or PCI Express device.
*/
if (pci_find_extcap(dev, PCIY_EXPRESS, &reg) == 0) {
/*
* Found a PCI Express capabilities register, this
* must be a PCI Express device.
*/
Use capability pointer to access PCIe registers rather than directly access them at fixed address. While I'm here don't touch other bits of PCIe device control register except max payload size. Reviewed by: marius
2009-11-22 19:11:34 +00:00
sc->bge_flags |= BGE_FLAG_PCIE;
sc->bge_expcap = reg;
Use pci_get_max_read_req() and pci_set_max_read_req() to set maximim read request size.
2010-03-25 17:17:35 +00:00
if (pci_get_max_read_req(dev) != 4096)
pci_set_max_read_req(dev, 4096);
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 {
/*
* Check if the device is in PCI-X Mode.
* (This bit is not valid on PCI Express controllers.)
*/
Use capability pointer to access PCIe registers rather than directly access them at fixed address. While I'm here don't touch other bits of PCIe device control register except max payload size. Reviewed by: marius
2009-11-22 19:11:34 +00:00
if (pci_find_extcap(dev, PCIY_PCIX, &reg) == 0)
sc->bge_pcixcap = reg;
- Limit BCM5701 B5 to 32-bit mode as a workaround for a bug which causes data corruption in combination with certain bridges. Information about this problem was kindly provided by davidch. [1] - As BGE_FLAG_PCIX is meant to indicate that the controller is in PCI-X mode, revert to the pre __FreeBSD_version 602101 method of reading the bus mode register rather than checking the mere existence of a PCI-X capability, which is also there when the NIC f.e. is put into a 32-bit slot causing it not to be in PCI-X mode. Setting BGE_FLAG_PCIX inappropriately could cause the NIC to be tuned incorrectly. PR: 128833 [1] Reviewed by: jhb MFC after: 3 days
2008-12-09 21:34:22 +00:00
if ((pci_read_config(dev, BGE_PCI_PCISTATE, 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
BGE_PCISTATE_PCI_BUSMODE) == 0)
sc->bge_flags |= BGE_FLAG_PCIX;
}
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_FLAG_40BIT_BUG should be set before creating DMA tags. Pointy hat to: yongari
2009-11-24 17:46:58 +00:00
/*
* The 40bit DMA bug applies to the 5714/5715 controllers and is
* not actually a MAC controller bug but an issue with the embedded
* PCIe to PCI-X bridge in the device. Use 40bit DMA workaround.
*/
if (BGE_IS_5714_FAMILY(sc) && (sc->bge_flags & BGE_FLAG_PCIX))
sc->bge_flags |= BGE_FLAG_40BIT_BUG;
Due to newly added PCIe capabilities fallback code for finding the PCIe capability did not work right on recent controllers. Remove FreeBSD 6.x support code. Reviewed by: marius
2009-11-22 18:47:56 +00:00
/*
* Allocate the interrupt, using MSI if possible. These devices
* support 8 MSI messages, but only the first one is used in
* normal operation.
*/
Use capability pointer to access PCIe registers rather than directly access them at fixed address. While I'm here don't touch other bits of PCIe device control register except max payload size. Reviewed by: marius
2009-11-22 19:11:34 +00:00
rid = 0;
Fix typo which inversed the logic which in turn disabled MSI. Pointy hat to: yongari
2009-11-25 17:51:14 +00:00
if (pci_find_extcap(sc->bge_dev, PCIY_MSI, &reg) == 0) {
Use capability pointer to access PCIe registers rather than directly access them at fixed address. While I'm here don't touch other bits of PCIe device control register except max payload size. Reviewed by: marius
2009-11-22 19:11:34 +00:00
sc->bge_msicap = reg;
if (bge_can_use_msi(sc)) {
msicount = pci_msi_count(dev);
if (msicount > 1)
msicount = 1;
} else
msicount = 0;
if (msicount == 1 && pci_alloc_msi(dev, &msicount) == 0) {
rid = 1;
sc->bge_flags |= BGE_FLAG_MSI;
}
}
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
sc->bge_irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &rid,
RF_SHAREABLE | RF_ACTIVE);
if (sc->bge_irq == NULL) {
device_printf(sc->bge_dev, "couldn't map interrupt\n");
error = ENXIO;
goto fail;
}
- In bge_ifmedia_upd_locked() take advantrage of LIST_FOREACH(). - If boot verbose, print asicrev, chiprev and bus type on attach. - For PCI Express devices: 1) Adjust max read request size to 4Kbytes 2) Turn on FIFO_LONG_BURST in RDMA during bge_blockinit() Though 1) does not seem to have much to do with the poor TX performance observed on PCI Express bge(4), 2) does fix the problem. [1] - Nuke the RX CPU self-diag, which prevents working cards from working (Linux tg3 does not have this diag neither does OpenBSD's bge(4)). The increasing of the firmware handshaking timeout to 20000 retries done as part of the original commit isn't merged as way already have a way higher BGE_TIMEOUT of 100000. PR: 119361 [1] Obtained from: tg3 via DragonflyBSD [1], DragonflyBSD
2009-03-21 00:23:07 +00:00
if (bootverbose)
device_printf(dev,
"CHIP ID 0x%08x; ASIC REV 0x%02x; CHIP REV 0x%02x; %s\n",
sc->bge_chipid, sc->bge_asicrev, sc->bge_chiprev,
(sc->bge_flags & BGE_FLAG_PCIX) ? "PCI-X" :
((sc->bge_flags & BGE_FLAG_PCIE) ? "PCI-E" : "PCI"));
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
BGE_LOCK_INIT(sc, device_get_nameunit(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
/* 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");
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;
Use new handshake command for BCM5750 or new controllers.
2010-01-22 18:46:37 +00:00
if (BGE_IS_575X_PLUS(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
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");
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
error = ENXIO;
goto fail;
}
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
error = bge_get_eaddr(sc, eaddr);
if (error) {
device_printf(sc->bge_dev,
"failed to read station address\n");
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
}
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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;
Split common parent DMA tag into ring DMA tag and TX/RX mbuf DMA tag. All controllers that are not BCM5755 or higher have 4GB boundary DMA bug. Previously bge(4) used 32bit DMA address to workaround the bug(r199670). However this caused the use of bounce buffers such that it resulted in poor performance for systems which have more than 4GB memory. Because bus_dma(9) honors boundary restriction requirement of DMA tag for dynamic buffers, having a separate TX/RX mbuf DMA tag will greatly reduce the possibility of using bounce buffers. For DMA buffers allocated with bus_dmamem_alloc(9), now bge(4) explicitly checks whether the requested memory region crossed the boundary or not. With this change, only the DMA buffer that crossed the boundary will use 32bit DMA address. Other DMA buffers are not affected as separate DMA tag is created for each DMA buffer. Even if 32bit DMA address space is used for a buffer, the chance to use bounce buffer is still very low as the size of buffer is small. This change should eliminate most usage of bounce buffers on systems that have more than 4GB memory. More correct fix would be teaching bus_dma(9) to honor boundary restriction for buffers created with bus_dmamem_alloc(9) but it seems that is not easy. While I'm here cleanup bge_dma_map_addr() and remove unnecessary member variables in bge_dmamap_arg structure. Tested by: marcel
2010-08-31 17:33:48 +00:00
if (bge_dma_alloc(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,
"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
error = ENXIO;
goto fail;
}
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
bge_add_sysctls(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 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
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
/* Initialize checksum features to use. */
sc->bge_csum_features = BGE_CSUM_FEATURES;
if (sc->bge_forced_udpcsum != 0)
sc->bge_csum_features |= CSUM_UDP;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
/* 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
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;
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);
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
ifp->if_hwassist = sc->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 |
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
IFCAP_VLAN_MTU;
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
if ((sc->bge_flags & BGE_FLAG_TSO) != 0) {
ifp->if_hwassist |= CSUM_TSO;
Add TSO support on VLAN. Controller requires VLAN hardware tagging to make TSO work on VLAN. So if VLAN hardware tagging is disabled explicitly clear TSO on VLAN. While I'm here remove duplicated VLAN_CAPABILITIES call.
2010-02-20 23:21:06 +00:00
ifp->if_capabilities |= IFCAP_TSO4 | IFCAP_VLAN_HWTSO;
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
}
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
#ifdef IFCAP_VLAN_HWCSUM
ifp->if_capabilities |= IFCAP_VLAN_HWCSUM;
#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
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;
Correct disabling checksum offloading for BCM5700 B0.
2009-11-09 00:16:50 +00:00
ifp->if_capenable &= ~IFCAP_HWCSUM;
- 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_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);
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
else if ((sc->bge_flags & BGE_FLAG_EADDR) &&
(sc->bge_asicrev != BGE_ASICREV_BCM5906)) {
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
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);
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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. */
For controllers that has dual mode PHY(copper or fiber) interfaces over GMII, make sure to enable GMII. With this change brgphy(4) is used to handle the dual mode PHY. Since we still don't have a sane way to pass PHY specific information to mii(4) layer special handling is needed in brgphy(4) to determine which mode of PHY was configured in parent interface. This change make BCM5715S work. Tested by: olli Obtained from: OpenBSD MFC after: 1 week
2010-01-14 19:08:43 +00:00
if ((pci_read_config(dev, BGE_PCI_SUBSYS, 4) >> 16) ==
SK_SUBSYSID_9D41 || (hwcfg & BGE_HWCFG_MEDIA) == BGE_MEDIA_FIBER) {
if (BGE_IS_5714_FAMILY(sc))
sc->bge_flags |= BGE_FLAG_MII_SERDES;
else
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) {
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +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,
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
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
*/
Don't reset the PHY probe retry counter within the loop so it will eventually terminate as intended. Submitted by: Helen Zhang
2009-02-10 21:54:23 +00:00
trys = 0;
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);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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");
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
bge_miibus_writereg(sc->bge_dev, 1, MII_BMCR,
BMCR_RESET);
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
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
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
Tell upper layer we support long frames. ether_ifattach() initializes it to ETHER_HDR_LEN so we have to override it after calling ether_ifattch(). While I'm here remove setting if_mtu value, it's initialized in ether_ifattach().
2009-11-07 20:37:38 +00:00
/* Tell upper layer we support long frames. */
ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
/*
* Hookup IRQ last.
*/
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
#if __FreeBSD_version > 700030
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
if (BGE_IS_5755_PLUS(sc) && sc->bge_flags & BGE_FLAG_MSI) {
/* Take advantage of single-shot MSI. */
Make sure one shot MSI is enabled. Submitted by: marius
2009-11-25 17:30:38 +00:00
CSR_WRITE_4(sc, BGE_MSI_MODE, CSR_READ_4(sc, BGE_MSI_MODE) &
~BGE_MSIMODE_ONE_SHOT_DISABLE);
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
sc->bge_tq = taskqueue_create_fast("bge_taskq", M_WAITOK,
taskqueue_thread_enqueue, &sc->bge_tq);
if (sc->bge_tq == NULL) {
device_printf(dev, "could not create taskqueue.\n");
ether_ifdetach(ifp);
error = ENXIO;
goto fail;
}
taskqueue_start_threads(&sc->bge_tq, 1, PI_NET, "%s taskq",
device_get_nameunit(sc->bge_dev));
error = bus_setup_intr(dev, sc->bge_irq,
INTR_TYPE_NET | INTR_MPSAFE, bge_msi_intr, NULL, sc,
&sc->bge_intrhand);
if (error)
ether_ifdetach(ifp);
} else
error = bus_setup_intr(dev, sc->bge_irq,
INTR_TYPE_NET | INTR_MPSAFE, NULL, bge_intr, sc,
&sc->bge_intrhand);
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
#else
error = bus_setup_intr(dev, sc->bge_irq, INTR_TYPE_NET | INTR_MPSAFE,
bge_intr, sc, &sc->bge_intrhand);
#endif
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
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
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
return (0);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
fail:
- Add support/quirks for the on-board BGEs found in Sun Blade 1500 Blade 2500, Fire V210 and probably some other sparc64 machines. These chips are typically not fitted with an EEPROM which means that we have to obtain the MAC address via OFW and that some chip tests will just always fail. These changes are based on the respective code found in OpenBSD with some additional info obtained from OpenSolaris and some style suggestions by jkim@. They also have the desired side-effect of respecting the 'local-mac-address?' system configuration variable for the affected BGEs. - In bge_attach() factor out calling bge_release_resources() before going to the fail label into the fail label as well as replace a magic 6 with ETHER_ADDR_LEN. Reviewed by: yongari (before style changes), jkim
2007-04-30 21:55:27 +00:00
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 (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);
- Add missing callout_drain() call. - Synchronize bge_tick() with callout_reset/callout_stop() calls. - Avoid using bge_tick() inside bge_link_upd(), use mii_pollstat() instead. MFC after: 2 month
2006-12-19 08:57:46 +00:00
callout_drain(&sc->bge_stat_ch);
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
if (sc->bge_tq)
taskqueue_drain(sc->bge_tq, &sc->bge_intr_task);
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
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
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
if (sc->bge_tq != NULL)
taskqueue_free(sc->bge_tq);
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 MSI support to the bge driver. I tested this on a Dell SC1435 running an i386 kernel, and it worked fine.
2006-12-14 23:10:12 +00:00
bus_release_resource(dev, SYS_RES_IRQ,
sc->bge_flags & BGE_FLAG_MSI ? 1 : 0, sc->bge_irq);
if (sc->bge_flags & BGE_FLAG_MSI)
pci_release_msi(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
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,
Prefer PCIR_BAR macro over BGE_PCI_BAR0.
2010-07-13 19:45:40 +00:00
PCIR_BAR(0), sc->bge_res);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 "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;
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
uint32_t cachesize, command, pcistate, reset, val;
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);
Use capability pointer to access PCIe registers rather than directly access them at fixed address. While I'm here don't touch other bits of PCIe device control register except max payload size. Reviewed by: marius
2009-11-22 19:11:34 +00:00
uint16_t devctl;
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +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
dev = sc->bge_dev;
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
if (BGE_IS_575X_PLUS(sc) && !BGE_IS_5714_FAMILY(sc) &&
(sc->bge_asicrev != BGE_ASICREV_BCM5906)) {
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)
write_op = bge_writemem_direct;
else
write_op = bge_writemem_ind;
Fix recent style problems and expand a comment. Submitted by: Bruce Evans
2006-12-13 21:13:09 +00:00
} else
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 = 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,
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
BGE_PCIMISCCTL_INDIRECT_ACCESS | BGE_PCIMISCCTL_MASK_PCI_INTR |
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 ||
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
BGE_IS_5755_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
if (bootverbose)
style.
2010-07-13 19:39:51 +00:00
device_printf(dev, "Disabling fastboot\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
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);
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
reset = BGE_MISCCFG_RESET_CORE_CLOCKS | BGE_32BITTIME_66MHZ;
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) {
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
if (CSR_READ_4(sc, 0x7E2C) == 0x60) /* PCIE 1.0 */
CSR_WRITE_4(sc, 0x7E2C, 0x20);
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) {
/* Prevent PCIE link training during global reset */
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
CSR_WRITE_4(sc, BGE_MISC_CFG, 1 << 29);
reset |= 1 << 29;
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
}
}
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Remove extra white space.
2009-11-22 18:30:19 +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))
Use Miscellaneous Configuration Register bit definition instead of magic number.
2010-07-15 23:34:58 +00:00
reset |= BGE_MISCCFG_GPHY_PD_OVERRIDE;
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
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
val = CSR_READ_4(sc, BGE_VCPU_STATUS);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
CSR_WRITE_4(sc, BGE_VCPU_STATUS,
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
val | BGE_VCPU_STATUS_DRV_RESET);
val = CSR_READ_4(sc, BGE_VCPU_EXT_CTRL);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
CSR_WRITE_4(sc, BGE_VCPU_EXT_CTRL,
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
val & ~BGE_VCPU_EXT_CTRL_HALT_CPU);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +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
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) {
DELAY(500000); /* wait for link training to complete */
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
val = pci_read_config(dev, 0xC4, 4);
pci_write_config(dev, 0xC4, val | (1 << 15), 4);
Add support for the BCM5750/5751. Unfortunately the documentation I have from Broadcom does not give much information on these devices, so the Broadcom Linux driver was used for clues to what these chips support. It turns out they are similar to the 5705 with the 5751 being the PCI-Express version and needing special work-arounds and settings.
2004-09-24 22:24:33 +00:00
}
Use capability pointer to access PCIe registers rather than directly access them at fixed address. While I'm here don't touch other bits of PCIe device control register except max payload size. Reviewed by: marius
2009-11-22 19:11:34 +00:00
devctl = pci_read_config(dev,
sc->bge_expcap + PCIR_EXPRESS_DEVICE_CTL, 2);
/* Clear enable no snoop and disable relaxed ordering. */
PCI express device status register has W1C feature. Writing 0 has no effect. Make sure to clear error bits by writing 1. [1] While I'm here use predefined value instead of hardcodig magic vlaue. Submitted by: msaitoh at NetBSD [1]
2010-02-01 20:58:45 +00:00
devctl &= ~(PCIM_EXP_CTL_RELAXED_ORD_ENABLE |
PCIM_EXP_CTL_NOSNOOP_ENABLE);
Use capability pointer to access PCIe registers rather than directly access them at fixed address. While I'm here don't touch other bits of PCIe device control register except max payload size. Reviewed by: marius
2009-11-22 19:11:34 +00:00
/* Set PCIE max payload size to 128. */
devctl &= ~PCIM_EXP_CTL_MAX_PAYLOAD;
pci_write_config(dev, sc->bge_expcap + PCIR_EXPRESS_DEVICE_CTL,
devctl, 2);
/* Clear error status. */
pci_write_config(dev, sc->bge_expcap + PCIR_EXPRESS_DEVICE_STA,
PCI express device status register has W1C feature. Writing 0 has no effect. Make sure to clear error bits by writing 1. [1] While I'm here use predefined value instead of hardcodig magic vlaue. Submitted by: msaitoh at NetBSD [1]
2010-02-01 20:58:45 +00:00
PCIM_EXP_STA_CORRECTABLE_ERROR |
PCIM_EXP_STA_NON_FATAL_ERROR | PCIM_EXP_STA_FATAL_ERROR |
PCIM_EXP_STA_UNSUPPORTED_REQ, 2);
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
}
- 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,
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
BGE_PCIMISCCTL_INDIRECT_ACCESS | BGE_PCIMISCCTL_MASK_PCI_INTR |
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 more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
write_op(sc, BGE_MISC_CFG, BGE_32BITTIME_66MHZ);
Set maximum read byte count to 2048 for PCI-X BCM5703/5704 devices. Also disable relaxed ordering as recommended by data sheet for PCI-X devices. For PCI-X BCM5704, set maximum outstanding split transactions to 0 as indicated by data sheet. For BCM5703 in PCI-X mode, DMA read watermark should be less than or equal to maximum read byte count configuration. Enforce this limitation in DMA read watermark configuration.
2010-03-10 20:54:08 +00:00
/*
* Disable PCI-X relaxed ordering to ensure status block update
Fix typo in r204978. Pointed out by: marius
2010-03-10 21:37:19 +00:00
* comes first then packet buffer DMA. Otherwise driver may
Set maximum read byte count to 2048 for PCI-X BCM5703/5704 devices. Also disable relaxed ordering as recommended by data sheet for PCI-X devices. For PCI-X BCM5704, set maximum outstanding split transactions to 0 as indicated by data sheet. For BCM5703 in PCI-X mode, DMA read watermark should be less than or equal to maximum read byte count configuration. Enforce this limitation in DMA read watermark configuration.
2010-03-10 20:54:08 +00:00
* read stale status block.
*/
if (sc->bge_flags & BGE_FLAG_PCIX) {
devctl = pci_read_config(dev,
sc->bge_pcixcap + PCIXR_COMMAND, 2);
devctl &= ~PCIXM_COMMAND_ERO;
if (sc->bge_asicrev == BGE_ASICREV_BCM5703) {
devctl &= ~PCIXM_COMMAND_MAX_READ;
devctl |= PCIXM_COMMAND_MAX_READ_2048;
} else if (sc->bge_asicrev == BGE_ASICREV_BCM5704) {
devctl &= ~(PCIXM_COMMAND_MAX_SPLITS |
PCIXM_COMMAND_MAX_READ);
devctl |= PCIXM_COMMAND_MAX_READ_2048;
}
pci_write_config(dev, sc->bge_pcixcap + PCIXR_COMMAND,
devctl, 2);
}
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
/* Re-enable MSI, if neccesary, and enable the 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)) {
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
/* This chip disables MSI on reset. */
if (sc->bge_flags & BGE_FLAG_MSI) {
Use capability pointer to access PCIe registers rather than directly access them at fixed address. While I'm here don't touch other bits of PCIe device control register except max payload size. Reviewed by: marius
2009-11-22 19:11:34 +00:00
val = pci_read_config(dev,
sc->bge_msicap + PCIR_MSI_CTRL, 2);
pci_write_config(dev,
sc->bge_msicap + PCIR_MSI_CTRL,
Re-enable MSI support for those chips on which it is believed to work properly.
2006-12-22 02:59:58 +00:00
val | PCIM_MSICTRL_MSI_ENABLE, 2);
val = CSR_READ_4(sc, BGE_MSI_MODE);
CSR_WRITE_4(sc, BGE_MSI_MODE,
val | BGE_MSIMODE_ENABLE);
}
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
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 support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5906) {
for (i = 0; i < BGE_TIMEOUT; i++) {
val = CSR_READ_4(sc, BGE_VCPU_STATUS);
if (val & BGE_VCPU_STATUS_INIT_DONE)
break;
DELAY(100);
}
if (i == BGE_TIMEOUT) {
style.
2010-07-13 19:39:51 +00:00
device_printf(dev, "reset timed out\n");
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
return (1);
}
} else {
/*
* Poll until we see the 1's complement of the magic number.
* This indicates that the firmware initialization is complete.
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
* We expect this to fail if no chip containing the Ethernet
* address is fitted though.
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
*/
for (i = 0; i < BGE_TIMEOUT; i++) {
DELAY(10);
val = bge_readmem_ind(sc, BGE_SOFTWARE_GENCOMM);
if (val == ~BGE_MAGIC_NUMBER)
break;
}
Whitespace cleanup.
2004-10-30 14:54:51 +00:00
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
if ((sc->bge_flags & BGE_FLAG_EADDR) && i == BGE_TIMEOUT)
style.
2010-07-13 19:39:51 +00:00
device_printf(dev,
"firmware handshake timed out, found 0x%08x\n",
val);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +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 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);
}
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Fix up byte swapping. */
Fix style(9) and consistency.
2007-03-06 20:14:48 +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) {
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
val = CSR_READ_4(sc, BGE_SERDES_CFG);
val = (val & ~0xFFF) | 0x880;
CSR_WRITE_4(sc, BGE_SERDES_CFG, val);
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
}
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 &&
Remove enabling Data FIFO protection with indirect memory access. r165114 added that code and that change ignored the same logic committed in r135772. In addition, data FIFO protection should be selectively enabled instead of applying to all PCIe devices. While I'm here add BCM5785 to devices that do not require this fix.
2010-07-14 21:47:49 +00:00
sc->bge_chipid != BGE_CHIPID_BCM5750_A0 &&
sc->bge_asicrev != BGE_ASICREV_BCM5785) {
/* Enable Data FIFO protection. */
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
val = CSR_READ_4(sc, 0x7C00);
CSR_WRITE_4(sc, 0x7C00, val | (1 << 25));
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
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
style.
2010-07-13 19:39:51 +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
}
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
static __inline void
bge_rxreuse_std(struct bge_softc *sc, int i)
{
struct bge_rx_bd *r;
r = &sc->bge_ldata.bge_rx_std_ring[sc->bge_std];
r->bge_flags = BGE_RXBDFLAG_END;
r->bge_len = sc->bge_cdata.bge_rx_std_seglen[i];
r->bge_idx = i;
BGE_INC(sc->bge_std, BGE_STD_RX_RING_CNT);
}
static __inline void
bge_rxreuse_jumbo(struct bge_softc *sc, int i)
{
struct bge_extrx_bd *r;
r = &sc->bge_ldata.bge_rx_jumbo_ring[sc->bge_jumbo];
r->bge_flags = BGE_RXBDFLAG_JUMBO_RING | BGE_RXBDFLAG_END;
r->bge_len0 = sc->bge_cdata.bge_rx_jumbo_seglen[i][0];
r->bge_len1 = sc->bge_cdata.bge_rx_jumbo_seglen[i][1];
r->bge_len2 = sc->bge_cdata.bge_rx_jumbo_seglen[i][2];
r->bge_len3 = sc->bge_cdata.bge_rx_jumbo_seglen[i][3];
r->bge_idx = i;
BGE_INC(sc->bge_jumbo, BGE_JUMBO_RX_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
/*
* 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
*/
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
bge_rxeof(struct bge_softc *sc, uint16_t rx_prod, int holdlck)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int rx_npkts = 0, stdcnt = 0, jumbocnt = 0;
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
uint16_t rx_cons;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 try to reevaluate current RX production index on each loop iteration as it can be updated by the card while we process the RX ring forcing us to process RX descriptors for which DMA synchronisation operation has not been performed. This fixes the bug when bge(4) drops packets under high load. Discussed with: yongari, marius Approved by: re (kib) MFC after: 1 week
2009-08-18 21:07:39 +00:00
rx_cons = sc->bge_rx_saved_considx;
o add locking o mark isr MPSAFE Supported by: FreeBSD Foundation
2003-11-11 17:57:03 +00:00
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
/* Nothing to do. */
- Do not try to reevaluate current RX production index on each loop iteration as it can be updated by the card while we process the RX ring forcing us to process RX descriptors for which DMA synchronisation operation has not been performed. This fixes the bug when bge(4) drops packets under high load. Discussed with: yongari, marius Approved by: re (kib) MFC after: 1 week
2009-08-18 21:07:39 +00:00
if (rx_cons == rx_prod)
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
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
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,
- On entrance to the rx_eof sync RX rings maps with POSTWRITE flag instead of POSTREAD: the hardware do not touch this memory (CPU updates it). It is already synchronized as PREWRITE after the processing is done. - Synchronize RX return ring memory in rx_eof. This is needed as the deviced updates this memory when receives packets. - Decouple the synchronization of BGE status block in the interrupt service routine: perfrom PREREAD synchronization only all accesses to this block are finished. This seems to be more natural. Reviewed by: yongari, marius MFC after: 2 weeks
2009-10-21 11:50:18 +00:00
sc->bge_cdata.bge_rx_std_ring_map, BUS_DMASYNC_POSTWRITE);
Do bus_dmamap_sync call only if frame size is greater than standard buffer size. If controller is not capable of handling jumbo frame, interface MTU couldn't be larger than standard MTU which in turn the received should be fit in standard buffer. This fixes bus_dmamap_sync call for jumbo ring is called even if interface is configured to use standard MTU. Also if total frame size could be fit into standard buffer don't use jumbo buffers.
2009-11-06 23:49:20 +00:00
if (ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN >
(MCLBYTES - ETHER_ALIGN))
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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,
- On entrance to the rx_eof sync RX rings maps with POSTWRITE flag instead of POSTREAD: the hardware do not touch this memory (CPU updates it). It is already synchronized as PREWRITE after the processing is done. - Synchronize RX return ring memory in rx_eof. This is needed as the deviced updates this memory when receives packets. - Decouple the synchronization of BGE status block in the interrupt service routine: perfrom PREREAD synchronization only all accesses to this block are finished. This seems to be more natural. Reviewed by: yongari, marius MFC after: 2 weeks
2009-10-21 11:50:18 +00:00
sc->bge_cdata.bge_rx_jumbo_ring_map, 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
- Do not try to reevaluate current RX production index on each loop iteration as it can be updated by the card while we process the RX ring forcing us to process RX descriptors for which DMA synchronisation operation has not been performed. This fixes the bug when bge(4) drops packets under high load. Discussed with: yongari, marius Approved by: re (kib) MFC after: 1 week
2009-08-18 21:07:39 +00:00
while (rx_cons != rx_prod) {
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
- Do not try to reevaluate current RX production index on each loop iteration as it can be updated by the card while we process the RX ring forcing us to process RX descriptors for which DMA synchronisation operation has not been performed. This fixes the bug when bge(4) drops packets under high load. Discussed with: yongari, marius Approved by: re (kib) MFC after: 1 week
2009-08-18 21:07:39 +00:00
cur_rx = &sc->bge_ldata.bge_rx_return_ring[rx_cons];
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
- Do not try to reevaluate current RX production index on each loop iteration as it can be updated by the card while we process the RX ring forcing us to process RX descriptors for which DMA synchronisation operation has not been performed. This fixes the bug when bge(4) drops packets under high load. Discussed with: yongari, marius Approved by: re (kib) MFC after: 1 week
2009-08-18 21:07:39 +00:00
BGE_INC(rx_cons, 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
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING &&
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) {
jumbocnt++;
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
m = sc->bge_cdata.bge_rx_jumbo_chain[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
if (cur_rx->bge_flags & BGE_RXBDFLAG_ERROR) {
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
bge_rxreuse_jumbo(sc, 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
continue;
}
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
if (bge_newbuf_jumbo(sc, rxidx) != 0) {
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
bge_rxreuse_jumbo(sc, rxidx);
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
ifp->if_iqdrops++;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
continue;
}
Fix I mssied in r199011. Rx ring index also should be updated. If we fill Rx ring full instead of half we can simplify this logic but this requires more experimentation.
2009-11-07 02:10:59 +00:00
BGE_INC(sc->bge_jumbo, BGE_JUMBO_RX_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
} else {
stdcnt++;
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
m = sc->bge_cdata.bge_rx_std_chain[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
if (cur_rx->bge_flags & BGE_RXBDFLAG_ERROR) {
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
bge_rxreuse_std(sc, 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
continue;
}
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
if (bge_newbuf_std(sc, rxidx) != 0) {
Fix a bug introduced in r199011. When bge(4) reuses loaded RX buffers it should also reinitialize RX descriptors otherwise some stale data could be passed to controller. This could end up with mbuf double free or unexpected NULL pointer dereference in upper stack. To fix the issue, save loaded buffer's length and reinitialize RX descriptors with the saved value whenever bge(4) reuses the loaded RX buffers. While I'm here, increase the number of RX buffers to 512 from 256. This simplifies RX buffer handling as well as giving more RX buffers. Controller supports just fixed number of RX buffers (i.e. 512) and bge(4) used to rely on hope that our CPU is fast enough to keep up with the controller. With this change, bge(4) will use 1MB for RX buffers but I don't think it would cause problems in these days. Reported by: marcel Tested by: marcel
2010-06-05 23:29:24 +00:00
bge_rxreuse_std(sc, rxidx);
Reimplement Rx buffer allocation to handle dma map load failure. Introduce two spare dma maps for standard buffer and jumbo buffer respectively. If loading a dma map failed reuse previously loaded dma map. This should fix unloaded dma map is used in case of dma map load failure. Also don't blindly unload dma map and defer dma map sync and unloading operation until we know dma map for new buffer is successfully loaded. This change saves unnecessary dma load/unload operation. Previously bge(4) tried to reuse mbuf with unloaded dma map which is really bad thing in bus_dma(9) perspective. While I'm here update if_iqdrops if we can't allocate Rx buffers.
2009-11-07 01:01:33 +00:00
ifp->if_iqdrops++;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
continue;
}
Fix I mssied in r199011. Rx ring index also should be updated. If we fill Rx ring full instead of half we can simplify this logic but this requires more experimentation.
2009-11-07 02:10:59 +00:00
BGE_INC(sc->bge_std, BGE_STD_RX_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
}
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;
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
if ((cur_rx->bge_ip_csum ^ 0xFFFF) == 0)
m->m_pkthdr.csum_flags |= CSUM_IP_VALID;
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
}
- 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) {
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
#if __FreeBSD_version > 700022
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;
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
#else
VLAN_INPUT_TAG_NEW(ifp, m, vlan_tag);
if (m == NULL)
continue;
#endif
- 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
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
if (holdlck != 0) {
BGE_UNLOCK(sc);
(*ifp->if_input)(ifp, m);
BGE_LOCK(sc);
} else
(*ifp->if_input)(ifp, m);
s/rk_npkts/rx_npkts Reported by: zec
2009-05-30 17:25:14 +00:00
rx_npkts++;
Try to workaround a race where bge_stop() may sneak in when bge_rxeof() drops and re-grabs the softc mutex in the middle, resulting in kernel trap 12. This may happen when a lot of traffic is being hammered on one bge(4) interface while the system is shutting down. Reported by: Alexander Sack <pisymbol gmail com> PR: kern/134548 MFC After: 2 weeks
2009-05-14 22:33:37 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING))
Fix return values appropriately. Tested by: zec
2009-05-30 17:56:19 +00:00
return (rx_npkts);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. Still need to fix autonegotiation support for 1000baseSX NICs, but beyond that, driver is pretty solid.
2001-09-27 23:55:28 +00:00
}
- On entrance to the rx_eof sync RX rings maps with POSTWRITE flag instead of POSTREAD: the hardware do not touch this memory (CPU updates it). It is already synchronized as PREWRITE after the processing is done. - Synchronize RX return ring memory in rx_eof. This is needed as the deviced updates this memory when receives packets. - Decouple the synchronization of BGE status block in the interrupt service routine: perfrom PREREAD synchronization only all accesses to this block are finished. This seems to be more natural. Reviewed by: yongari, marius MFC after: 2 weeks
2009-10-21 11:50:18 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_rx_return_ring_tag,
sc->bge_cdata.bge_rx_return_ring_map, BUS_DMASYNC_PREREAD);
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
Do bus_dmamap_sync call only if frame size is greater than standard buffer size. If controller is not capable of handling jumbo frame, interface MTU couldn't be larger than standard MTU which in turn the received should be fit in standard buffer. This fixes bus_dmamap_sync call for jumbo ring is called even if interface is configured to use standard MTU. Also if total frame size could be fit into standard buffer don't use jumbo buffers.
2009-11-06 23:49:20 +00:00
if (jumbocnt > 0)
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
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
- Do not try to reevaluate current RX production index on each loop iteration as it can be updated by the card while we process the RX ring forcing us to process RX descriptors for which DMA synchronisation operation has not been performed. This fixes the bug when bge(4) drops packets under high load. Discussed with: yongari, marius Approved by: re (kib) MFC after: 1 week
2009-08-18 21:07:39 +00:00
sc->bge_rx_saved_considx = rx_cons;
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_RX_CONS0_LO, 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
if (stdcnt)
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_RX_STD_PROD_LO, sc->bge_std);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 (jumbocnt)
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_RX_JUMBO_PROD_LO, sc->bge_jumbo);
Partially revert r199035. Revision 1.158 says only lower ten bits of BGE_RXLP_LOCSTAT_IFIN_DROPS register is valid. For BCM5761 case it seems the controller maintains 16bits value for the register. However 16bits are still too small to count all dropped packets happened in a second. To get a correct counter we have to read the register in bge_rxeof() which would be too expensive. Pointed out by: bde
2009-11-08 19:59:54 +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
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
bge_txeof(struct bge_softc *sc, uint16_t tx_cons)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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() already checks whether it has to free transmitted mbufs or not by comparing reported TX consumer index with saved index. So remove unnecessary check done after freeing transmitted mbufs. While I'm here nuke unnecessary variable initializations.
2010-08-31 19:59:18 +00:00
struct bge_tx_bd *cur_tx;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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;
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. */
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
if (sc->bge_tx_saved_considx == tx_cons)
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
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,
Controller does not update Tx descriptors(send BDs) after sending frames so remove unnecessary BUS_DMASYNC_PREREAD and BUS_DMASYNC_POSTREAD of bus_dmamap_sync(9).
2009-11-10 20:29:20 +00:00
sc->bge_cdata.bge_tx_ring_map, 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.
*/
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
while (sc->bge_tx_saved_considx != tx_cons) {
bge_txeof() already checks whether it has to free transmitted mbufs or not by comparing reported TX consumer index with saved index. So remove unnecessary check done after freeing transmitted mbufs. While I'm here nuke unnecessary variable initializations.
2010-08-31 19:59:18 +00:00
uint32_t 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
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) {
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_tx_mtag,
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_tx_dmamap[idx],
BUS_DMASYNC_POSTWRITE);
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_tx_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
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);
}
bge_txeof() already checks whether it has to free transmitted mbufs or not by comparing reported TX consumer index with saved index. So remove unnecessary check done after freeing transmitted mbufs. While I'm here nuke unnecessary variable initializations.
2010-08-31 19:59:18 +00:00
ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
In bge_txeof(), cancel the watchdog timeout if all descriptors have been handled instead of when at least one descriptor was just handled. For bge, it is normal to get a txeof when only a small fraction of the queued tx descriptors have been handled, so the bug broke the watchdog in a usual case.
2006-12-20 12:03:21 +00:00
if (sc->bge_txcnt == 0)
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
}
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
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
static int
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
bge_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
{
struct bge_softc *sc = ifp->if_softc;
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
uint16_t rx_prod, tx_cons;
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
uint32_t statusword;
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
int rx_npkts = 0;
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);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
- style(9) cleanup. - Fix comments and printf()s about allocating jumbo buffers.
2006-06-07 21:03:20 +00:00
}
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,
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
sc->bge_cdata.bge_status_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
rx_prod = sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx;
tx_cons = sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx;
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
Remove unnecessary atomic operation in bge_poll. bge(4) always holds a driver lock in the function entry and memory synchronization is handled by bus_dmamap_sync(9).
2010-08-31 20:56:18 +00:00
statusword = sc->bge_ldata.bge_status_block->bge_status;
sc->bge_ldata.bge_status_block->bge_status = 0;
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,
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
sc->bge_cdata.bge_status_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
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
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
/* Note link event. It will be processed by POLL_AND_CHECK_STATUS. */
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 (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;
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
rx_npkts = bge_rxeof(sc, rx_prod, 1);
Try to workaround a race where bge_stop() may sneak in when bge_rxeof() drops and re-grabs the softc mutex in the middle, resulting in kernel trap 12. This may happen when a lot of traffic is being hammered on one bge(4) interface while the system is shutting down. Reported by: Alexander Sack <pisymbol gmail com> PR: kern/134548 MFC After: 2 weeks
2009-05-14 22:33:37 +00:00
if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
BGE_UNLOCK(sc);
Fix return values appropriately. Tested by: zec
2009-05-30 17:56:19 +00:00
return (rx_npkts);
Try to workaround a race where bge_stop() may sneak in when bge_rxeof() drops and re-grabs the softc mutex in the middle, resulting in kernel trap 12. This may happen when a lot of traffic is being hammered on one bge(4) interface while the system is shutting down. Reported by: Alexander Sack <pisymbol gmail com> PR: kern/134548 MFC After: 2 weeks
2009-05-14 22:33:37 +00:00
}
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
bge_txeof(sc, tx_cons);
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 (!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);
When user_frac in the polling subsystem is low it is going to busy the CPU for too long period than necessary. Additively, interfaces are kept polled (in the tick) even if no more packets are available. In order to avoid such situations a new generic mechanism can be implemented in proactive way, keeping track of the time spent on any packet and fragmenting the time for any tick, stopping the processing as soon as possible. In order to implement such mechanism, the polling handler needs to change, returning the number of packets processed. While the intended logic is not part of this patch, the polling KPI is broken by this commit, adding an int return value and the new flag IFCAP_POLLING_NOCOUNT (which will signal that the return value is meaningless for the installed handler and checking should be skipped). Bump __FreeBSD_version in order to signal such situation. Reviewed by: emaste Sponsored by: Sandvine Incorporated
2009-05-30 15:14:44 +00:00
return (rx_npkts);
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 */
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
static int
bge_msi_intr(void *arg)
{
struct bge_softc *sc;
sc = (struct bge_softc *)arg;
/*
* This interrupt is not shared and controller already
* disabled further interrupt.
*/
taskqueue_enqueue(sc->bge_tq, &sc->bge_intr_task);
return (FILTER_HANDLED);
}
static void
bge_intr_task(void *arg, int pending)
{
struct bge_softc *sc;
struct ifnet *ifp;
uint32_t status;
uint16_t rx_prod, tx_cons;
sc = (struct bge_softc *)arg;
ifp = sc->bge_ifp;
if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
return;
/* Get updated status block. */
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map,
BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
/* Save producer/consumer indexess. */
rx_prod = sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx;
tx_cons = sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx;
status = sc->bge_ldata.bge_status_block->bge_status;
sc->bge_ldata.bge_status_block->bge_status = 0;
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
/* Let controller work. */
bge_writembx(sc, BGE_MBX_IRQ0_LO, 0);
if ((status & BGE_STATFLAG_LINKSTATE_CHANGED) != 0) {
BGE_LOCK(sc);
bge_link_upd(sc);
BGE_UNLOCK(sc);
}
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
/* Check RX return ring producer/consumer. */
bge_rxeof(sc, rx_prod, 0);
}
if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
BGE_LOCK(sc);
/* Check TX ring producer/consumer. */
bge_txeof(sc, tx_cons);
if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
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
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;
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
uint16_t rx_prod, tx_cons;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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
Avoid a race and a pessimization in bge_intr(): - moved the synchronizing bus read to after the bus write for the first interrupt ack so that it actually synchronizes everything necessary. We were acking not only the status update that triggered the interrupt together with any status updates that occurred before we got around to the bus write for the ack, but also any status updates that occur after we do the bus write but before the write reaches the device. The corresponding race for the second interrupt ack resulted in sometimes returning from the interrupt handler with acked but unserviced interrupt events. Such events then remain unserviced until further events cause another interrupt or the watchdog times out. The race was often lost on my 5705, apparently since my 5705 has broken event coalescing which causes a status update for almost every packet, so another status update is quite likely to occur while the interrupt handler is running. Watchdog timeouts weren't very noticeable, apparently because bge_txeof() has one of the usual bugs resetting the watchdog. - don't disable device interrupts while bge_intr() is running. Doing this just had the side effects of: - entering a device mode in which different coalescing parameters apply. Different coalescing parameters can be used to either inhibit or enhance the chance of getting another status update while in the interrupt handler. This feature is useless with the current organization of the interrupt handler but might be useful with a taskqueue handler. - giving a race for ack+reenable/return. This cannot be handled by simply rearranging the order of bus accesses like the race for ack+keepenable/entry. It is necessary to sync the ack and then check for new events. - taking longer, especially with the extra code to avoid the race on ack+reenable/return. Reviewed by: ru, gleb, scottl
2006-12-20 11:14:45 +00:00
/*
* Ack the interrupt by writing something to BGE_MBX_IRQ0_LO. Don't
* disable interrupts by writing nonzero like we used to, since with
* our current organization this just gives complications and
* pessimizations for re-enabling interrupts. We used to have races
* instead of the necessary complications. Disabling interrupts
* would just reduce the chance of a status update while we are
* running (by switching to the interrupt-mode coalescence
* parameters), but this chance is already very low so it is more
* efficient to get another interrupt than prevent it.
*
* We do the ack first to ensure another interrupt if there is a
* status update after the ack. We don't check for the status
* changing later because it is more efficient to get another
* interrupt than prevent it, not quite as above (not checking is
* a smaller optimization than not toggling the interrupt enable,
* since checking doesn't involve PCI accesses and toggling require
* the status check). So toggling would probably be a pessimization
* even with MSI. It would only be needed for using a task queue.
*/
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_IRQ0_LO, 0);
Avoid a race and a pessimization in bge_intr(): - moved the synchronizing bus read to after the bus write for the first interrupt ack so that it actually synchronizes everything necessary. We were acking not only the status update that triggered the interrupt together with any status updates that occurred before we got around to the bus write for the ack, but also any status updates that occur after we do the bus write but before the write reaches the device. The corresponding race for the second interrupt ack resulted in sometimes returning from the interrupt handler with acked but unserviced interrupt events. Such events then remain unserviced until further events cause another interrupt or the watchdog times out. The race was often lost on my 5705, apparently since my 5705 has broken event coalescing which causes a status update for almost every packet, so another status update is quite likely to occur while the interrupt handler is running. Watchdog timeouts weren't very noticeable, apparently because bge_txeof() has one of the usual bugs resetting the watchdog. - don't disable device interrupts while bge_intr() is running. Doing this just had the side effects of: - entering a device mode in which different coalescing parameters apply. Different coalescing parameters can be used to either inhibit or enhance the chance of getting another status update while in the interrupt handler. This feature is useless with the current organization of the interrupt handler but might be useful with a taskqueue handler. - giving a race for ack+reenable/return. This cannot be handled by simply rearranging the order of bus accesses like the race for ack+keepenable/entry. It is necessary to sync the ack and then check for new events. - taking longer, especially with the extra code to avoid the race on ack+reenable/return. Reviewed by: ru, gleb, scottl
2006-12-20 11:14:45 +00:00
Revert r205090. It's hard to know when the mail box register write will get flushed to the hardware and it may take longer. Pointed out by: scottl
2010-03-16 17:45:16 +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;
/* 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 | BUS_DMASYNC_POSTWRITE);
rx_prod = sc->bge_ldata.bge_status_block->bge_idx[0].bge_rx_prod_idx;
tx_cons = sc->bge_ldata.bge_status_block->bge_idx[0].bge_tx_cons_idx;
sc->bge_ldata.bge_status_block->bge_status = 0;
bus_dmamap_sync(sc->bge_cdata.bge_status_tag,
sc->bge_cdata.bge_status_map,
BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
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. */
For MSI case, interrupt is not shared and we don't need to force PCI flush to get correct status block update. Add an optimized interrupt handler that is activated for MSI case. Actual interrupt handling is done by taskqueue such that the handler does not require driver lock for Rx path. The MSI capable bge(4) controllers automatically disables further interrupt once it enters interrupt state so we don't need PIO access to disable interrupt in interrupt handler.
2009-11-22 20:31:40 +00:00
bge_rxeof(sc, rx_prod, 1);
Try to workaround a race where bge_stop() may sneak in when bge_rxeof() drops and re-grabs the softc mutex in the middle, resulting in kernel trap 12. This may happen when a lot of traffic is being hammered on one bge(4) interface while the system is shutting down. Reported by: Alexander Sack <pisymbol gmail com> PR: kern/134548 MFC After: 2 weeks
2009-05-14 22:33:37 +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 workaround a race where bge_stop() may sneak in when bge_rxeof() drops and re-grabs the softc mutex in the middle, resulting in kernel trap 12. This may happen when a lot of traffic is being hammered on one bge(4) interface while the system is shutting down. Reported by: Alexander Sack <pisymbol gmail com> PR: kern/134548 MFC After: 2 weeks
2009-05-14 22:33:37 +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 TX ring producer/consumer. */
Cache Rx producer/Tx consumer index as soon as we know status block update and then clear status block. Previously it used to access these index without synchronization which may cause problems when bounce buffers are used. Also add missing bus_dmamap_sync(9) in polling handler. Since we now update status block in driver, adjust bus_dmamap_sync(9) for status block.
2009-11-22 20:02:13 +00:00
bge_txeof(sc, tx_cons);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 &&
!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 {
Fix a long standing ASF heartbeat sending bug. The initial implementation of heartbeat interval was 2 but there was typo which caused the heartbeat is sent approximately every 5 seconds. This caused unintended controller reset by firmware because firmware thought OS was crashed. Submitted by: Floris Bos < info <> je-eigen-domein dot nl > Tested by: Andrzej Tobola < ato <> iem dot pw dot edu dot pl >
2010-01-22 18:35:50 +00:00
sc->bge_asf_count = 2;
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_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,
Use correct bits to stop firmware when ASF mode is enabled.
2007-03-08 00:49:26 +00:00
CSR_READ_4(sc, BGE_CPU_EVENT) | (1 << 14));
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
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 missing callout_drain() call. - Synchronize bge_tick() with callout_reset/callout_stop() calls. - Avoid using bge_tick() inside bge_link_upd(), use mii_pollstat() instead. MFC after: 2 month
2006-12-19 08:57:46 +00:00
/* Synchronize with possible callout reset/stop. */
if (callout_pending(&sc->bge_stat_ch) ||
!callout_active(&sc->bge_stat_ch))
return;
- 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);
bge_tick(): do not touch PHY if link is up. This should solve problem with extra input errors for some BCM57XX chips. PR: kern/122295
2008-09-08 18:10:15 +00:00
/*
* Do not touch PHY if we have link up. This could break
* IPMI/ASF mode or produce extra input errors
* (extra errors was reported for bcm5701 & bcm5704).
*/
if (!sc->bge_link)
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
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++;
Use a better approach to force the interrupt which should work for all cards/modes. In addition to the intr forcing added with rev. 1.205 adopt the other places to use the same logic. We need to exclude a few chips/revisions (5700, 5788) from using the enhanced version and fall back to the old way as that is the only method they support. Tested by: phk Suggested by: davidch, Broadcom (thanks a lot for the help!) MFC after: 16 days
2008-05-05 18:42:17 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 ||
sc->bge_flags & BGE_FLAG_5788)
BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_INTR_SET);
else
BGE_SETBIT(sc, BGE_HCC_MODE, BGE_HCCMODE_COAL_NOW);
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
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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));
Count number of inbound packets which were chosen to be discarded as input errors. Also count out of receive BDs as input errors.
2009-11-08 01:30:35 +00:00
ifp->if_ierrors += CSR_READ_4(sc, BGE_RXLP_LOCSTAT_OUT_OF_BDS);
- 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);
Count number of inbound packets which were chosen to be discarded as input errors. Also count out of receive BDs as input errors.
2009-11-08 01:30:35 +00:00
ifp->if_ierrors += CSR_READ_4(sc, BGE_RXLP_LOCSTAT_IFIN_ERRORS);
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;
Fix style(9). Pointed out by: many
2007-02-12 23:33:05 +00:00
#define READ_STAT(sc, stats, stat) \
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_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
Fix style(9). Pointed out by: many
2007-02-12 23:33:05 +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;
}
}
Remove extra white space.
2009-11-22 18:30:19 +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
/* 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);
}
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
static struct mbuf *
bge_setup_tso(struct bge_softc *sc, struct mbuf *m, uint16_t *mss)
{
struct ip *ip;
struct tcphdr *tcp;
struct mbuf *n;
uint16_t hlen;
Remove Tx mbuf parsing code for VLAN in TSO path. Controller does not support TSO over VLAN if VLAN hardware tagging is disabled so there is no need to check VLAN here. While I'm here make sure to pullup IP/TCP headers in the first buffer.
2010-02-22 21:03:15 +00:00
uint32_t poff;
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
if (M_WRITABLE(m) == 0) {
/* Get a writable copy. */
n = m_dup(m, M_DONTWAIT);
m_freem(m);
if (n == NULL)
return (NULL);
m = n;
}
Remove Tx mbuf parsing code for VLAN in TSO path. Controller does not support TSO over VLAN if VLAN hardware tagging is disabled so there is no need to check VLAN here. While I'm here make sure to pullup IP/TCP headers in the first buffer.
2010-02-22 21:03:15 +00:00
m = m_pullup(m, sizeof(struct ether_header) + sizeof(struct ip));
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
if (m == NULL)
return (NULL);
Remove Tx mbuf parsing code for VLAN in TSO path. Controller does not support TSO over VLAN if VLAN hardware tagging is disabled so there is no need to check VLAN here. While I'm here make sure to pullup IP/TCP headers in the first buffer.
2010-02-22 21:03:15 +00:00
ip = (struct ip *)(mtod(m, char *) + sizeof(struct ether_header));
poff = sizeof(struct ether_header) + (ip->ip_hl << 2);
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
m = m_pullup(m, poff + sizeof(struct tcphdr));
if (m == NULL)
return (NULL);
tcp = (struct tcphdr *)(mtod(m, char *) + poff);
Remove Tx mbuf parsing code for VLAN in TSO path. Controller does not support TSO over VLAN if VLAN hardware tagging is disabled so there is no need to check VLAN here. While I'm here make sure to pullup IP/TCP headers in the first buffer.
2010-02-22 21:03:15 +00:00
m = m_pullup(m, poff + (tcp->th_off << 2));
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
if (m == NULL)
return (NULL);
/*
* It seems controller doesn't modify IP length and TCP pseudo
* checksum. These checksum computed by upper stack should be 0.
*/
*mss = m->m_pkthdr.tso_segsz;
ip->ip_sum = 0;
ip->ip_len = htons(*mss + (ip->ip_hl << 2) + (tcp->th_off << 2));
/* Clear pseudo checksum computed by TCP stack. */
tcp->th_sum = 0;
/*
* Broadcom controllers uses different descriptor format for
* TSO depending on ASIC revision. Due to TSO-capable firmware
* license issue and lower performance of firmware based TSO
* we only support hardware based TSO which is applicable for
* BCM5755 or newer controllers. Hardware based TSO uses 11
* bits to store MSS and upper 5 bits are used to store IP/TCP
* header length(including IP/TCP options). The header length
* is expressed as 32 bits unit.
*/
hlen = ((ip->ip_hl << 2) + (tcp->th_off << 2)) >> 2;
*mss |= (hlen << 11);
return (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
/*
* 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;
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
uint16_t csum_flags, mss, vlan_tag;
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;
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
mss = 0;
vlan_tag = 0;
if ((m->m_pkthdr.csum_flags & CSUM_TSO) != 0) {
*m_head = m = bge_setup_tso(sc, m, &mss);
if (*m_head == NULL)
return (ENOBUFS);
csum_flags |= BGE_TXBDFLAG_CPU_PRE_DMA |
BGE_TXBDFLAG_CPU_POST_DMA;
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
} else if ((m->m_pkthdr.csum_flags & sc->bge_csum_features) != 0) {
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
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;
}
Add workaround to overcome hardware limitation which allows only a single outstanding DMA read operation. Most controllers targeted to client with PCIe bus interface(e.g. BCM5761) may have this limitation. All controllers for servers does not have this limitation. Collapsing mbuf chains to reduce number of memory reads before transmitting was most effective way to workaround this. I got about 940Mbps from 850Mbps with mbuf collapsing on BCM5761. However it takes a lot of CPU cycles to collapse mbuf chains so add tunable to control the number of allowed TX buffers before collapsing. The default value is 0 which effectively disables the forced collapsing. For most cases 2 would yield best performance(about 930Mbps) without much sacrificing CPU cycles. Note the collapsing is only activated when the controller is on PCIe bus and the frame does not need TSO operation. TSO does not seem to suffer from the hardware limitation because the payload size is much bigger than normal IP datagram. Thanks to davidch@ who told me the limitation of client controllers and actually gave possible workarounds to mitigate the limitation. Reviewed by: davidch, marius
2009-12-03 23:57:06 +00:00
if ((m->m_pkthdr.csum_flags & CSUM_TSO) == 0 &&
Create sysctl node(dev.bge.%d.focred_collapse) instead of hw.bge.forced_collapse. hw.bge.forced_collapse affects all bge(4) controllers on system which may not desirable behavior of the sysctl node. Also allow the sysctl node could be modified at any time. Reviewed by: bde (initial version)
2009-12-08 17:54:23 +00:00
sc->bge_forced_collapse > 0 &&
(sc->bge_flags & BGE_FLAG_PCIE) != 0 && m->m_next != NULL) {
Add workaround to overcome hardware limitation which allows only a single outstanding DMA read operation. Most controllers targeted to client with PCIe bus interface(e.g. BCM5761) may have this limitation. All controllers for servers does not have this limitation. Collapsing mbuf chains to reduce number of memory reads before transmitting was most effective way to workaround this. I got about 940Mbps from 850Mbps with mbuf collapsing on BCM5761. However it takes a lot of CPU cycles to collapse mbuf chains so add tunable to control the number of allowed TX buffers before collapsing. The default value is 0 which effectively disables the forced collapsing. For most cases 2 would yield best performance(about 930Mbps) without much sacrificing CPU cycles. Note the collapsing is only activated when the controller is on PCIe bus and the frame does not need TSO operation. TSO does not seem to suffer from the hardware limitation because the payload size is much bigger than normal IP datagram. Thanks to davidch@ who told me the limitation of client controllers and actually gave possible workarounds to mitigate the limitation. Reviewed by: davidch, marius
2009-12-03 23:57:06 +00:00
/*
* Forcedly collapse mbuf chains to overcome hardware
* limitation which only support a single outstanding
* DMA read operation.
*/
Create sysctl node(dev.bge.%d.focred_collapse) instead of hw.bge.forced_collapse. hw.bge.forced_collapse affects all bge(4) controllers on system which may not desirable behavior of the sysctl node. Also allow the sysctl node could be modified at any time. Reviewed by: bde (initial version)
2009-12-08 17:54:23 +00:00
if (sc->bge_forced_collapse == 1)
Add workaround to overcome hardware limitation which allows only a single outstanding DMA read operation. Most controllers targeted to client with PCIe bus interface(e.g. BCM5761) may have this limitation. All controllers for servers does not have this limitation. Collapsing mbuf chains to reduce number of memory reads before transmitting was most effective way to workaround this. I got about 940Mbps from 850Mbps with mbuf collapsing on BCM5761. However it takes a lot of CPU cycles to collapse mbuf chains so add tunable to control the number of allowed TX buffers before collapsing. The default value is 0 which effectively disables the forced collapsing. For most cases 2 would yield best performance(about 930Mbps) without much sacrificing CPU cycles. Note the collapsing is only activated when the controller is on PCIe bus and the frame does not need TSO operation. TSO does not seem to suffer from the hardware limitation because the payload size is much bigger than normal IP datagram. Thanks to davidch@ who told me the limitation of client controllers and actually gave possible workarounds to mitigate the limitation. Reviewed by: davidch, marius
2009-12-03 23:57:06 +00:00
m = m_defrag(m, M_DONTWAIT);
else
Create sysctl node(dev.bge.%d.focred_collapse) instead of hw.bge.forced_collapse. hw.bge.forced_collapse affects all bge(4) controllers on system which may not desirable behavior of the sysctl node. Also allow the sysctl node could be modified at any time. Reviewed by: bde (initial version)
2009-12-08 17:54:23 +00:00
m = m_collapse(m, M_DONTWAIT, sc->bge_forced_collapse);
Don't free mbuf chains when bge(4) fails to collapse the mbuf chains. This part of code is to enhance performance so failing the collapsing should not free TX frames. Otherwise bge(4) will unnecessarily drop frames which in turn can freeze the network connection. Reported by: Igor Sysoev (is <> rambler-co dot ru) Tested by: Igor Sysoev (is <> rambler-co dot ru)
2010-01-15 17:55:18 +00:00
if (m == NULL)
m = *m_head;
Add workaround to overcome hardware limitation which allows only a single outstanding DMA read operation. Most controllers targeted to client with PCIe bus interface(e.g. BCM5761) may have this limitation. All controllers for servers does not have this limitation. Collapsing mbuf chains to reduce number of memory reads before transmitting was most effective way to workaround this. I got about 940Mbps from 850Mbps with mbuf collapsing on BCM5761. However it takes a lot of CPU cycles to collapse mbuf chains so add tunable to control the number of allowed TX buffers before collapsing. The default value is 0 which effectively disables the forced collapsing. For most cases 2 would yield best performance(about 930Mbps) without much sacrificing CPU cycles. Note the collapsing is only activated when the controller is on PCIe bus and the frame does not need TSO operation. TSO does not seem to suffer from the hardware limitation because the payload size is much bigger than normal IP datagram. Thanks to davidch@ who told me the limitation of client controllers and actually gave possible workarounds to mitigate the limitation. Reviewed by: davidch, marius
2009-12-03 23:57:06 +00:00
*m_head = 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
map = sc->bge_cdata.bge_tx_dmamap[idx];
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_tx_mtag, map, m, segs,
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
&nsegs, BUS_DMA_NOWAIT);
if (error == EFBIG) {
- Use more appropriate maxsize, nsegments and maxsegsize parameters when creating the parent bus DMA tag. While at it correct the style and a nearby comment. - Take advantage of m_collapse(9) for performance reasons. MFC after: 2 weeks
2008-04-26 10:54:17 +00:00
m = m_collapse(m, M_DONTWAIT, BGE_NSEG_NEW);
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 (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;
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
error = bus_dmamap_load_mbuf_sg(sc->bge_cdata.bge_tx_mtag, map,
m, segs, &nsegs, BUS_DMA_NOWAIT);
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 (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
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
/* Check if we have enough free send BDs. */
if (sc->bge_txcnt + nsegs >= BGE_TX_RING_CNT) {
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_unload(sc->bge_cdata.bge_tx_mtag, map);
o Rewrite bge_encap() to use bus_dmamap_load_mbuf_sg(9), inlining the callback function bge_dma_map_tx_desc() into the bge_encap() itself. o If busdma returns EFBIG, try to m_defrag() the packet. Reviewed by: yongari
2005-12-15 09:45:53 +00:00
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
Remove common DMA tag used for TX/RX mbufs and create Tx DMA tag and Rx DMA tag separately. Previously it used a common mbuf DMA tag for both Tx and Rx path but Rx buffer(standard ring case) should have a single DMA segment and maximum buffer size of the segment should be less than or equal to MCLBYTES. This change also make it possible to add TSO with minor changes.
2009-11-04 20:57:52 +00:00
bus_dmamap_sync(sc->bge_cdata.bge_tx_mtag, map, BUS_DMASYNC_PREWRITE);
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
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
#if __FreeBSD_version > 700022
if (m->m_flags & M_VLANTAG) {
csum_flags |= BGE_TXBDFLAG_VLAN_TAG;
vlan_tag = m->m_pkthdr.ether_vtag;
}
#else
{
struct m_tag *mtag;
if ((mtag = VLAN_OUTPUT_TAG(sc->bge_ifp, m)) != NULL) {
csum_flags |= BGE_TXBDFLAG_VLAN_TAG;
vlan_tag = VLAN_TAG_VALUE(mtag);
}
}
#endif
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;
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
d->bge_vlan_tag = vlan_tag;
d->bge_mss = mss;
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 (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
Convert bge(4) to use busdma. I have not tested this on anything besides x86 yet since 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;
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
struct mbuf *m_head;
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;
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
int 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
sc = ifp->if_softc;
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +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
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
if (!sc->bge_link ||
(ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
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;
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
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
for (count = 0; !IFQ_DRV_IS_EMPTY(&ifp->if_snd);) {
if (sc->bge_txcnt > BGE_TX_RING_CNT - 16) {
ifp->if_drv_flags |= IFF_DRV_OACTIVE;
break;
}
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.
*/
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
#ifdef ETHER_BPF_MTAP
Partially back out rev. 1.148 and use new ETHER_BPF_MTAP() macro for VLAN. Remaining changes are cosmetic.
2006-12-20 01:12:07 +00:00
ETHER_BPF_MTAP(ifp, m_head);
Pollute bge(4) with #if's and #ifdef's to make MFC easier. MFC after: 3 days
2007-03-06 19:15:16 +00:00
#else
BPF_MTAP(ifp, m_head);
#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
}
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
if (count > 0) {
bus_dmamap_sync(sc->bge_cdata.bge_tx_ring_tag,
sc->bge_cdata.bge_tx_ring_map, BUS_DMASYNC_PREWRITE);
/* Transmit. */
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_TX_HOST_PROD0_LO, prodidx);
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
/* 5700 b2 errata */
if (sc->bge_chiprev == BGE_CHIPREV_5700_BX)
bge_writembx(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
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
sc->bge_tx_prodidx = prodidx;
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
Rearrange bge_start_locked to see we can send more frames by checking IFF_DRV_RUNNING and IFF_DRV_OACTIVE flags. Also if we have less than 16 free send BDs set IFF_DRV_OACTIVE and try it later. Previously bge(4) used to reserve 16 free send BDs after loading dma maps but hardware just need one reserved send BD. If prouder index has the same value of consumer index it means the Tx queue is empty. While I'm here check IFQ_DRV_IS_EMPTY first to save one lock operation.
2009-11-22 19:44:11 +00:00
/*
* Set a timeout in case the chip goes out to lunch.
*/
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 +
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
ETHER_HDR_LEN + ETHER_CRC_LEN +
(ifp->if_capenable & IFCAP_VLAN_MTU ? ETHER_VLAN_ENCAP_LEN : 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
/* 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);
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
/* Program VLAN tag stripping. */
bge_setvlan(sc);
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
/* Override UDP checksum offloading. */
if (sc->bge_forced_udpcsum == 0)
sc->bge_csum_features &= ~CSUM_UDP;
else
sc->bge_csum_features |= CSUM_UDP;
if (ifp->if_capabilities & IFCAP_TXCSUM &&
ifp->if_capenable & IFCAP_TXCSUM) {
ifp->if_hwassist &= ~(BGE_CSUM_FEATURES | CSUM_UDP);
ifp->if_hwassist |= sc->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
/* Init RX ring. */
Make bge_newbuf_std()/bge_newbuf_jumbo() returns actual error code for buffer allocation. If driver know we are out of Rx buffers let controller stop. This should fix panic when interface is run even if it had no configured Rx buffers.
2009-11-04 21:06:54 +00:00
if (bge_init_rx_ring_std(sc) != 0) {
device_printf(sc->bge_dev, "no memory for std Rx buffers.\n");
bge_stop(sc);
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
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. */
Do bus_dmamap_sync call only if frame size is greater than standard buffer size. If controller is not capable of handling jumbo frame, interface MTU couldn't be larger than standard MTU which in turn the received should be fit in standard buffer. This fixes bus_dmamap_sync call for jumbo ring is called even if interface is configured to use standard MTU. Also if total frame size could be fit into standard buffer don't use jumbo buffers.
2009-11-06 23:49:20 +00:00
if (ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN >
(MCLBYTES - ETHER_ALIGN)) {
Make bge_newbuf_std()/bge_newbuf_jumbo() returns actual error code for buffer allocation. If driver know we are out of Rx buffers let controller stop. This should fix panic when interface is run even if it had no configured Rx buffers.
2009-11-04 21:06:54 +00:00
if (bge_init_rx_ring_jumbo(sc) != 0) {
style.
2010-07-13 19:39:51 +00:00
device_printf(sc->bge_dev,
Fix error message for jumbo buffer allocation failure.
2010-07-13 19:42:55 +00:00
"no memory for jumbo Rx buffers.\n");
Make bge_newbuf_std()/bge_newbuf_jumbo() returns actual error code for buffer allocation. If driver know we are out of Rx buffers let controller stop. This should fix panic when interface is run even if it had no configured Rx buffers.
2009-11-04 21:06:54 +00:00
bge_stop(sc);
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
- 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);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_IRQ0_LO, 1);
Introduce polling(4) capability for bge(4). Submitted by: Oleg Bulyzhin <oleg rinet.ru>
2005-10-22 14:31:01 +00:00
} 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);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(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
}
Remove extra white space.
2009-11-22 18:30:19 +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;
- In bge_ifmedia_upd_locked() take advantrage of LIST_FOREACH(). - If boot verbose, print asicrev, chiprev and bus type on attach. - For PCI Express devices: 1) Adjust max read request size to 4Kbytes 2) Turn on FIFO_LONG_BURST in RDMA during bge_blockinit() Though 1) does not seem to have much to do with the poor TX performance observed on PCI Express bge(4), 2) does fix the problem. [1] - Nuke the RX CPU self-diag, which prevents working cards from working (Linux tg3 does not have this diag neither does OpenBSD's bge(4)). The increasing of the firmware handshaking timeout to 20000 retries done as part of the original commit isn't merged as way already have a way higher BGE_TIMEOUT of 100000. PR: 119361 [1] Obtained from: tg3 via DragonflyBSD [1], DragonflyBSD
2009-03-21 00:23:07 +00:00
struct mii_softc *miisc;
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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 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.
*/
Remove BCM5704S specific tunable (hw.bge.fake_autoneg) and auto-detect the condition.
2007-05-22 19:35:34 +00:00
if (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;
Remove BCM5704S specific tunable (hw.bge.fake_autoneg) and auto-detect the condition.
2007-05-22 19:35:34 +00:00
sgdig = CSR_READ_4(sc, BGE_SGDIG_STS);
if (sgdig & BGE_SGDIGSTS_DONE) {
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);
}
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
}
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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);
- In bge_ifmedia_upd_locked() take advantrage of LIST_FOREACH(). - If boot verbose, print asicrev, chiprev and bus type on attach. - For PCI Express devices: 1) Adjust max read request size to 4Kbytes 2) Turn on FIFO_LONG_BURST in RDMA during bge_blockinit() Though 1) does not seem to have much to do with the poor TX performance observed on PCI Express bge(4), 2) does fix the problem. [1] - Nuke the RX CPU self-diag, which prevents working cards from working (Linux tg3 does not have this diag neither does OpenBSD's bge(4)). The increasing of the firmware handshaking timeout to 20000 retries done as part of the original commit isn't merged as way already have a way higher BGE_TIMEOUT of 100000. PR: 119361 [1] Obtained from: tg3 via DragonflyBSD [1], DragonflyBSD
2009-03-21 00:23:07 +00:00
if (mii->mii_instance)
LIST_FOREACH(miisc, &mii->mii_phys, mii_list)
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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_phy_reset(miisc);
mii_mediachg(mii);
In some situations we were not clearing pending link state attentions. Because of this we were not getting further interrupts for link state changes, thus never went into iface UP state and thus could not transmit. The only way out of this was an incoming packet generating an rx interrupt and making us call into bge_link_upd. Up to rev. 1.101, in bge_start_locked, we only returned instantly if there was 'no link AND nothing queued for tx'. So with a packet queued for tx, we hit the register scrubbing at the end of bge_start_locked and were out fine. We simply lost a packet or two but got the interrupts need to get into UP state. With rev. 1.102 this was turned into 'if there is no link OR there is nothing to send' (correct behaviour) and as long as there is no link we never hit the register scrubbing and consequently never got the link UP. What we do now is force an interrupt at the end of bge_ifmedia_upd_locked so we will call bge_link_upd, clear the link state attention and get further interrupts. This helps to get the iface UP on an idle network or at least to get it UP faster not depending on an rx intr anymore. In case you could not get a DHCP lease or it took very long, it was because of this. It is unknown which chips are affected by this. ASIC rev. 0x2003 was the most popular trouble candidate. At least the fiber cards should have been working fine. Which register to scrub is currently under discussion. The comitted solution was tested and found to work for a lot of setups. It might not help with MSI. The reason why we end up in such a situation is entirely unknown. PR: kern/111804 Tested by: phk, scottl at Y! MFC after: 14 days
2008-04-08 11:51:17 +00:00
/*
* Force an interrupt so that we will call bge_link_upd
* if needed and clear any pending link state attention.
* Without this we are not getting any further interrupts
* for link state changes and thus will not UP the link and
* not be able to send in bge_start_locked. The only
* way to get things working was to receive a packet and
* get an RX intr.
* bge_tick should help for fiber cards and we might not
* need to do this here if BGE_FLAG_TBI is set but as
* we poll for fiber anyway it should not harm.
*/
Use a better approach to force the interrupt which should work for all cards/modes. In addition to the intr forcing added with rev. 1.205 adopt the other places to use the same logic. We need to exclude a few chips/revisions (5700, 5788) from using the enhanced version and fall back to the old way as that is the only method they support. Tested by: phk Suggested by: davidch, Broadcom (thanks a lot for the help!) MFC after: 16 days
2008-05-05 18:42:17 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5700 ||
sc->bge_flags & BGE_FLAG_5788)
BGE_SETBIT(sc, BGE_MISC_LOCAL_CTL, BGE_MLC_INTR_SET);
else
BGE_SETBIT(sc, BGE_HCC_MODE, BGE_HCCMODE_COAL_NOW);
In some situations we were not clearing pending link state attentions. Because of this we were not getting further interrupts for link state changes, thus never went into iface UP state and thus could not transmit. The only way out of this was an incoming packet generating an rx interrupt and making us call into bge_link_upd. Up to rev. 1.101, in bge_start_locked, we only returned instantly if there was 'no link AND nothing queued for tx'. So with a packet queued for tx, we hit the register scrubbing at the end of bge_start_locked and were out fine. We simply lost a packet or two but got the interrupts need to get into UP state. With rev. 1.102 this was turned into 'if there is no link OR there is nothing to send' (correct behaviour) and as long as there is no link we never hit the register scrubbing and consequently never got the link UP. What we do now is force an interrupt at the end of bge_ifmedia_upd_locked so we will call bge_link_upd, clear the link state attention and get further interrupts. This helps to get the iface UP on an idle network or at least to get it UP faster not depending on an rx intr anymore. In case you could not get a DHCP lease or it took very long, it was because of this. It is unknown which chips are affected by this. ASIC rev. 0x2003 was the most popular trouble candidate. At least the fiber cards should have been working fine. Which register to scrub is currently under discussion. The comitted solution was tested and found to work for a lot of setups. It might not help with MSI. The reason why we end up in such a situation is entirely unknown. PR: kern/111804 Tested by: phk, scottl at Y! MFC after: 14 days
2008-04-08 11:51:17 +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
}
/*
* 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);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(sc, BGE_MBX_IRQ0_LO, 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);
BGE_CLRBIT(sc, BGE_PCI_MISC_CTL,
BGE_PCIMISCCTL_MASK_PCI_INTR);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_writembx(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
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)
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
ifp->if_hwassist |= sc->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
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
ifp->if_hwassist &= ~sc->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
}
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
Implement TSO for BCM5755 or newer controllers. Some controllers seem to require a special firmware to use TSO. But the firmware is not available to FreeBSD and Linux claims that the TSO performed by the firmware is slower than hardware based TSO. Moreover the firmware based TSO has one known bug which can't handle TSO if ethernet header + IP/TCP header is greater than 80 bytes. The workaround for the TSO bug exist but it seems it's too expensive than not using TSO at all. Some hardwares also have the TSO bug so limit the TSO to the controllers that are not affected TSO issues (e.g. 5755 or higher). While I'm here set VLAN tag bit to all descriptors that belengs to a frame instead of the first descriptor of a frame. The datasheet is not clear how to handle VLAN tag bit but it worked either way in my testing. This makes it simplify TSO configuration a little bit. Big thanks to davidch@ who sent me detailed TSO information. Without this I was not able to implement it. Tested by: current
2009-11-22 21:16:30 +00:00
if ((mask & IFCAP_TSO4) != 0 &&
(ifp->if_capabilities & IFCAP_TSO4) != 0) {
ifp->if_capenable ^= IFCAP_TSO4;
if ((ifp->if_capenable & IFCAP_TSO4) != 0)
ifp->if_hwassist |= CSUM_TSO;
else
ifp->if_hwassist &= ~CSUM_TSO;
}
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
if (mask & IFCAP_VLAN_MTU) {
ifp->if_capenable ^= IFCAP_VLAN_MTU;
ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
bge_init(sc);
}
Add TSO support on VLAN. Controller requires VLAN hardware tagging to make TSO work on VLAN. So if VLAN hardware tagging is disabled explicitly clear TSO on VLAN. While I'm here remove duplicated VLAN_CAPABILITIES call.
2010-02-20 23:21:06 +00:00
if ((mask & IFCAP_VLAN_HWTSO) != 0 &&
(ifp->if_capabilities & IFCAP_VLAN_HWTSO) != 0)
ifp->if_capenable ^= IFCAP_VLAN_HWTSO;
if ((mask & IFCAP_VLAN_HWTAGGING) != 0 &&
(ifp->if_capabilities & IFCAP_VLAN_HWTAGGING) != 0) {
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
Add TSO support on VLAN. Controller requires VLAN hardware tagging to make TSO work on VLAN. So if VLAN hardware tagging is disabled explicitly clear TSO on VLAN. While I'm here remove duplicated VLAN_CAPABILITIES call.
2010-02-20 23:21:06 +00:00
if ((ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0)
ifp->if_capenable &= ~IFCAP_VLAN_HWTSO;
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
BGE_LOCK(sc);
bge_setvlan(sc);
BGE_UNLOCK(sc);
Add TSO support on VLAN. Controller requires VLAN hardware tagging to make TSO work on VLAN. So if VLAN hardware tagging is disabled explicitly clear TSO on VLAN. While I'm here remove duplicated VLAN_CAPABILITIES call.
2010-02-20 23:21:06 +00:00
}
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +00:00
#ifdef VLAN_CAPABILITIES
Add TSO support on VLAN. Controller requires VLAN hardware tagging to make TSO work on VLAN. So if VLAN hardware tagging is disabled explicitly clear TSO on VLAN. While I'm here remove duplicated VLAN_CAPABILITIES call.
2010-02-20 23:21:06 +00:00
VLAN_CAPABILITIES(ifp);
Add on/off controls for VLAN_MTU and VLAN_HWTAGGING to bge(4).
2007-06-01 02:02:39 +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
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;
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
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
Immediately after clearing a pending callout that didn't make it due to the lock we hold, disable interrupts, and announce to the firmware that we are shutting down. Especially do this before disabling blocks. This makes some types of machines with asf enabled no longer hang upon boot, when we start configuring the interface. PR: i386/96382, kern/100410, kern/122252, kern/116328 Reported by: erwin Hardware provided by: TDC A/S Reviewed by: stas Tested by: stas
2009-10-13 20:22:12 +00:00
/* Disable host interrupts. */
BGE_SETBIT(sc, BGE_PCI_MISC_CTL, BGE_PCIMISCCTL_MASK_PCI_INTR);
bge_writembx(sc, BGE_MBX_IRQ0_LO, 1);
/*
* Tell firmware we're shutting down.
*/
bge_stop_fw(sc);
bge_sig_pre_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
/*
- 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);
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
CSR_WRITE_4(sc, BGE_FTQ_RESET, 0xFFFFFFFF);
Add device driver support for the Broadcom BCM570x family of gigabit ethernet controllers. This adds support for the 3Com 3c996-T, the SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on Dell PowerEdge 2550 servers. The latter configuration hauls ass: preliminary measurements show TCP speeds of over 900Mbps using only normal size frames. TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping are supported, as well as interrupt moderation. 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, 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
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_reset(sc);
bge_sig_legacy(sc, BGE_RESET_STOP);
bge_sig_post_reset(sc, BGE_RESET_STOP);
Remove extra white space.
2009-11-22 18:30:19 +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
* 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);
sc->bge_tx_saved_considx = BGE_TXCONS_UNSET;
- Add missing callout_drain() call. - Synchronize bge_tick() with callout_reset/callout_stop() calls. - Avoid using bge_tick() inside bge_link_upd(), use mii_pollstat() instead. MFC after: 2 month
2006-12-19 08:57:46 +00:00
/* Clear MAC's link state (PHY may still have link UP). */
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
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.
*/
device_shutdown returns an int.
2009-02-05 18:43:13 +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_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);
device_shutdown returns an int.
2009-02-05 18:43:13 +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
}
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) {
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);
- Add missing callout_drain() call. - Synchronize bge_tick() with callout_reset/callout_stop() calls. - Avoid using bge_tick() inside bge_link_upd(), use mii_pollstat() instead. MFC after: 2 month
2006-12-19 08:57:46 +00:00
mii_pollstat(mii);
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_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);
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
CSR_WRITE_4(sc, BGE_MAC_STS, 0xFFFFFFFF);
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 (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
} else if (CSR_READ_4(sc, BGE_MI_MODE) & BGE_MIMODE_AUTOPOLL) {
Whitespace.
2006-06-08 10:19:16 +00:00
/*
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +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.
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
*/
link = (CSR_READ_4(sc, BGE_MI_STS) & BGE_MISTS_LINK) ? 1 : 0;
if (link != sc->bge_link ||
sc->bge_asicrev == BGE_ASICREV_BCM5700) {
mii = device_get_softc(sc->bge_miibus);
- Add missing callout_drain() call. - Synchronize bge_tick() with callout_reset/callout_stop() calls. - Avoid using bge_tick() inside bge_link_upd(), use mii_pollstat() instead. MFC after: 2 month
2006-12-19 08:57:46 +00:00
mii_pollstat(mii);
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_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");
}
}
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
} else {
/*
* Discard link events for MII/GMII controllers
* if MI auto-polling is disabled.
*/
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. */
Fix style(9) and consistency.
2007-03-06 20:14:48 +00:00
CSR_WRITE_4(sc, BGE_MAC_STS, BGE_MACSTAT_SYNC_CHANGED |
BGE_MACSTAT_CFG_CHANGED | BGE_MACSTAT_MI_COMPLETE |
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_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
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
#define BGE_SYSCTL_STAT(sc, ctx, desc, parent, node, oid) \
Fix some OID names and minor style as per feedback from various people. Also, apparently quad support is broken in the sysctl infrastructure, so don't pretend that it works. Submitted by: ru, bde
2007-03-13 00:41:55 +00:00
SYSCTL_ADD_PROC(ctx, parent, OID_AUTO, oid, CTLTYPE_UINT|CTLFLAG_RD, \
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
sc, offsetof(struct bge_stats, node), bge_sysctl_stats, "IU", \
desc)
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;
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
struct sysctl_oid_list *children, *schildren;
struct sysctl_oid *tree;
Load tunable from loader.conf(5) instead of device.hints(5).
2010-08-21 23:13:16 +00:00
char tn[32];
int unit;
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
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");
#endif
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
Load tunable from loader.conf(5) instead of device.hints(5).
2010-08-21 23:13:16 +00:00
unit = device_get_unit(sc->bge_dev);
Create sysctl node(dev.bge.%d.focred_collapse) instead of hw.bge.forced_collapse. hw.bge.forced_collapse affects all bge(4) controllers on system which may not desirable behavior of the sysctl node. Also allow the sysctl node could be modified at any time. Reviewed by: bde (initial version)
2009-12-08 17:54:23 +00:00
/*
* A common design characteristic for many Broadcom client controllers
* is that they only support a single outstanding DMA read operation
* on the PCIe bus. This means that it will take twice as long to fetch
* a TX frame that is split into header and payload buffers as it does
* to fetch a single, contiguous TX frame (2 reads vs. 1 read). For
* these controllers, coalescing buffers to reduce the number of memory
* reads is effective way to get maximum performance(about 940Mbps).
* Without collapsing TX buffers the maximum TCP bulk transfer
* performance is about 850Mbps. However forcing coalescing mbufs
* consumes a lot of CPU cycles, so leave it off by default.
*/
Load tunable from loader.conf(5) instead of device.hints(5).
2010-08-21 23:13:16 +00:00
sc->bge_forced_collapse = 0;
snprintf(tn, sizeof(tn), "dev.bge.%d.forced_collapse", unit);
TUNABLE_INT_FETCH(tn, &sc->bge_forced_collapse);
Create sysctl node(dev.bge.%d.focred_collapse) instead of hw.bge.forced_collapse. hw.bge.forced_collapse affects all bge(4) controllers on system which may not desirable behavior of the sysctl node. Also allow the sysctl node could be modified at any time. Reviewed by: bde (initial version)
2009-12-08 17:54:23 +00:00
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "forced_collapse",
CTLFLAG_RW, &sc->bge_forced_collapse, 0,
"Number of fragmented TX buffers of a frame allowed before "
"forced collapsing");
It seems all Broadcom controllers have a bug that can generate UDP datagrams with checksum value 0 when TX UDP checksum offloading is enabled. Generating UDP checksum value 0 is RFC 768 violation. Even though the probability of generating such UDP datagrams is low, I don't want to see FreeBSD boxes to inject such datagrams into network so disable UDP checksum offloading by default. Users still override this behavior by setting a sysctl variable or loader tunable, dev.bge.%d.forced_udpcsum. I have no idea why this issue was not reported so far given that bge(4) is one of the most commonly used controller on high-end server class systems. Thanks to andre@ who passed the PR to me. PR: kern/104826
2010-08-22 01:39:09 +00:00
/*
* It seems all Broadcom controllers have a bug that can generate UDP
* datagrams with checksum value 0 when TX UDP checksum offloading is
* enabled. Generating UDP checksum value 0 is RFC 768 violation.
* Even though the probability of generating such UDP datagrams is
* low, I don't want to see FreeBSD boxes to inject such datagrams
* into network so disable UDP checksum offloading by default. Users
* still override this behavior by setting a sysctl variable,
* dev.bge.0.forced_udpcsum.
*/
sc->bge_forced_udpcsum = 0;
snprintf(tn, sizeof(tn), "dev.bge.%d.bge_forced_udpcsum", unit);
TUNABLE_INT_FETCH(tn, &sc->bge_forced_udpcsum);
SYSCTL_ADD_INT(ctx, children, OID_AUTO, "forced_udpcsum",
CTLFLAG_RW, &sc->bge_forced_udpcsum, 0,
"Enable UDP checksum offloading even if controller can "
"generate UDP checksum value 0");
Do not report MAC, TX, and RX stats via sysctl(8) with BCM5705+. The register layout is little different from memory-mapped stats in the previous generation chips. In fact, it is bad because registers in this range are cleared after reading them. Reviewed by: scottl MFC after: 3 days
2007-11-16 16:39:27 +00:00
if (BGE_IS_5705_PLUS(sc))
return;
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
tree = SYSCTL_ADD_NODE(ctx, children, OID_AUTO, "stats", CTLFLAG_RD,
NULL, "BGE Statistics");
schildren = children = SYSCTL_CHILDREN(tree);
BGE_SYSCTL_STAT(sc, ctx, "Frames Dropped Due To Filters",
children, COSFramesDroppedDueToFilters,
"FramesDroppedDueToFilters");
BGE_SYSCTL_STAT(sc, ctx, "NIC DMA Write Queue Full",
children, nicDmaWriteQueueFull, "DmaWriteQueueFull");
BGE_SYSCTL_STAT(sc, ctx, "NIC DMA Write High Priority Queue Full",
children, nicDmaWriteHighPriQueueFull, "DmaWriteHighPriQueueFull");
BGE_SYSCTL_STAT(sc, ctx, "NIC No More RX Buffer Descriptors",
children, nicNoMoreRxBDs, "NoMoreRxBDs");
Fix some OID names and minor style as per feedback from various people. Also, apparently quad support is broken in the sysctl infrastructure, so don't pretend that it works. Submitted by: ru, bde
2007-03-13 00:41:55 +00:00
BGE_SYSCTL_STAT(sc, ctx, "Discarded Input Frames",
children, ifInDiscards, "InputDiscards");
BGE_SYSCTL_STAT(sc, ctx, "Input Errors",
children, ifInErrors, "InputErrors");
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
BGE_SYSCTL_STAT(sc, ctx, "NIC Recv Threshold Hit",
children, nicRecvThresholdHit, "RecvThresholdHit");
BGE_SYSCTL_STAT(sc, ctx, "NIC DMA Read Queue Full",
children, nicDmaReadQueueFull, "DmaReadQueueFull");
BGE_SYSCTL_STAT(sc, ctx, "NIC DMA Read High Priority Queue Full",
children, nicDmaReadHighPriQueueFull, "DmaReadHighPriQueueFull");
BGE_SYSCTL_STAT(sc, ctx, "NIC Send Data Complete Queue Full",
children, nicSendDataCompQueueFull, "SendDataCompQueueFull");
BGE_SYSCTL_STAT(sc, ctx, "NIC Ring Set Send Producer Index",
children, nicRingSetSendProdIndex, "RingSetSendProdIndex");
BGE_SYSCTL_STAT(sc, ctx, "NIC Ring Status Update",
children, nicRingStatusUpdate, "RingStatusUpdate");
BGE_SYSCTL_STAT(sc, ctx, "NIC Interrupts",
children, nicInterrupts, "Interrupts");
BGE_SYSCTL_STAT(sc, ctx, "NIC Avoided Interrupts",
children, nicAvoidedInterrupts, "AvoidedInterrupts");
BGE_SYSCTL_STAT(sc, ctx, "NIC Send Threshold Hit",
children, nicSendThresholdHit, "SendThresholdHit");
tree = SYSCTL_ADD_NODE(ctx, schildren, OID_AUTO, "rx", CTLFLAG_RD,
NULL, "BGE RX Statistics");
children = SYSCTL_CHILDREN(tree);
BGE_SYSCTL_STAT(sc, ctx, "Inbound Octets",
children, rxstats.ifHCInOctets, "Octets");
BGE_SYSCTL_STAT(sc, ctx, "Fragments",
children, rxstats.etherStatsFragments, "Fragments");
BGE_SYSCTL_STAT(sc, ctx, "Inbound Unicast Packets",
children, rxstats.ifHCInUcastPkts, "UcastPkts");
BGE_SYSCTL_STAT(sc, ctx, "Inbound Multicast Packets",
children, rxstats.ifHCInMulticastPkts, "MulticastPkts");
BGE_SYSCTL_STAT(sc, ctx, "FCS Errors",
children, rxstats.dot3StatsFCSErrors, "FCSErrors");
BGE_SYSCTL_STAT(sc, ctx, "Alignment Errors",
children, rxstats.dot3StatsAlignmentErrors, "AlignmentErrors");
BGE_SYSCTL_STAT(sc, ctx, "XON Pause Frames Received",
children, rxstats.xonPauseFramesReceived, "xonPauseFramesReceived");
BGE_SYSCTL_STAT(sc, ctx, "XOFF Pause Frames Received",
children, rxstats.xoffPauseFramesReceived,
"xoffPauseFramesReceived");
BGE_SYSCTL_STAT(sc, ctx, "MAC Control Frames Received",
children, rxstats.macControlFramesReceived,
"ControlFramesReceived");
BGE_SYSCTL_STAT(sc, ctx, "XOFF State Entered",
children, rxstats.xoffStateEntered, "xoffStateEntered");
BGE_SYSCTL_STAT(sc, ctx, "Frames Too Long",
children, rxstats.dot3StatsFramesTooLong, "FramesTooLong");
BGE_SYSCTL_STAT(sc, ctx, "Jabbers",
children, rxstats.etherStatsJabbers, "Jabbers");
BGE_SYSCTL_STAT(sc, ctx, "Undersized Packets",
children, rxstats.etherStatsUndersizePkts, "UndersizePkts");
BGE_SYSCTL_STAT(sc, ctx, "Inbound Range Length Errors",
Fix some OID names and minor style as per feedback from various people. Also, apparently quad support is broken in the sysctl infrastructure, so don't pretend that it works. Submitted by: ru, bde
2007-03-13 00:41:55 +00:00
children, rxstats.inRangeLengthError, "inRangeLengthError");
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
BGE_SYSCTL_STAT(sc, ctx, "Outbound Range Length Errors",
Fix some OID names and minor style as per feedback from various people. Also, apparently quad support is broken in the sysctl infrastructure, so don't pretend that it works. Submitted by: ru, bde
2007-03-13 00:41:55 +00:00
children, rxstats.outRangeLengthError, "outRangeLengthError");
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
tree = SYSCTL_ADD_NODE(ctx, schildren, OID_AUTO, "tx", CTLFLAG_RD,
NULL, "BGE TX Statistics");
children = SYSCTL_CHILDREN(tree);
BGE_SYSCTL_STAT(sc, ctx, "Outbound Octets",
children, txstats.ifHCOutOctets, "Octets");
BGE_SYSCTL_STAT(sc, ctx, "TX Collisions",
children, txstats.etherStatsCollisions, "Collisions");
BGE_SYSCTL_STAT(sc, ctx, "XON Sent",
children, txstats.outXonSent, "XonSent");
BGE_SYSCTL_STAT(sc, ctx, "XOFF Sent",
children, txstats.outXoffSent, "XoffSent");
BGE_SYSCTL_STAT(sc, ctx, "Flow Control Done",
children, txstats.flowControlDone, "flowControlDone");
BGE_SYSCTL_STAT(sc, ctx, "Internal MAC TX errors",
children, txstats.dot3StatsInternalMacTransmitErrors,
"InternalMacTransmitErrors");
BGE_SYSCTL_STAT(sc, ctx, "Single Collision Frames",
children, txstats.dot3StatsSingleCollisionFrames,
"SingleCollisionFrames");
BGE_SYSCTL_STAT(sc, ctx, "Multiple Collision Frames",
children, txstats.dot3StatsMultipleCollisionFrames,
"MultipleCollisionFrames");
Remove extra white space.
2009-11-22 18:30:19 +00:00
BGE_SYSCTL_STAT(sc, ctx, "Deferred Transmissions",
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
children, txstats.dot3StatsDeferredTransmissions,
"DeferredTransmissions");
BGE_SYSCTL_STAT(sc, ctx, "Excessive Collisions",
children, txstats.dot3StatsExcessiveCollisions,
"ExcessiveCollisions");
BGE_SYSCTL_STAT(sc, ctx, "Late Collisions",
Fix some OID names and minor style as per feedback from various people. Also, apparently quad support is broken in the sysctl infrastructure, so don't pretend that it works. Submitted by: ru, bde
2007-03-13 00:41:55 +00:00
children, txstats.dot3StatsLateCollisions,
"LateCollisions");
Remove extra white space.
2009-11-22 18:30:19 +00:00
BGE_SYSCTL_STAT(sc, ctx, "Outbound Unicast Packets",
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
children, txstats.ifHCOutUcastPkts, "UcastPkts");
BGE_SYSCTL_STAT(sc, ctx, "Outbound Multicast Packets",
children, txstats.ifHCOutMulticastPkts, "MulticastPkts");
BGE_SYSCTL_STAT(sc, ctx, "Outbound Broadcast Packets",
children, txstats.ifHCOutBroadcastPkts, "BroadcastPkts");
BGE_SYSCTL_STAT(sc, ctx, "Carrier Sense Errors",
children, txstats.dot3StatsCarrierSenseErrors,
"CarrierSenseErrors");
BGE_SYSCTL_STAT(sc, ctx, "Outbound Discards",
children, txstats.ifOutDiscards, "Discards");
BGE_SYSCTL_STAT(sc, ctx, "Outbound Errors",
children, txstats.ifOutErrors, "Errors");
}
static int
bge_sysctl_stats(SYSCTL_HANDLER_ARGS)
{
struct bge_softc *sc;
Fix some OID names and minor style as per feedback from various people. Also, apparently quad support is broken in the sysctl infrastructure, so don't pretend that it works. Submitted by: ru, bde
2007-03-13 00:41:55 +00:00
uint32_t result;
Do not report MAC, TX, and RX stats via sysctl(8) with BCM5705+. The register layout is little different from memory-mapped stats in the previous generation chips. In fact, it is bad because registers in this range are cleared after reading them. Reviewed by: scottl MFC after: 3 days
2007-11-16 16:39:27 +00:00
int offset;
Add MAC, RX, and TX stats reporting via sysctl.
2007-03-12 09:25:57 +00:00
sc = (struct bge_softc *)arg1;
offset = arg2;
Do not report MAC, TX, and RX stats via sysctl(8) with BCM5705+. The register layout is little different from memory-mapped stats in the previous generation chips. In fact, it is bad because registers in this range are cleared after reading them. Reviewed by: scottl MFC after: 3 days
2007-11-16 16:39:27 +00:00
result = CSR_READ_4(sc, BGE_MEMWIN_START + BGE_STATS_BLOCK + offset +
offsetof(bge_hostaddr, bge_addr_lo));
Despite several examples in the kernel, the third argument of sysctl_handle_int is not sizeof the int type you want to export. The type must always be an int or an unsigned int. Remove the instances where a sizeof(variable) is passed to stop people accidently cut and pasting these examples. In a few places this was sysctl_handle_int was being used on 64 bit types, which would truncate the value to be exported. In these cases use sysctl_handle_quad to export them and change the format to Q so that sysctl(1) can still print them.
2007-06-04 18:25:08 +00:00
return (sysctl_handle_int(oidp, &result, 0, req));
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
}
#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");
Fix more style(9) bugs. - Remove some excessive parentheses around shift operators. - Use macro instead of magic number where it is applicable. - Change lower-case hexdecimals to upper cases to match wpaul's style. - Revert some unnecessary line wraps and changes from the previous commit. Pointed out by: bde
2007-03-08 00:29:18 +00:00
for (i = 0x800; i < 0xA00; ) {
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("%06x:", i);
for (j = 0; j < 8; j++) {
printf(" %08x", CSR_READ_4(sc, i));
i += 4;
}
printf("\n");
}
printf("Hardware Flags:\n");
- Add support for new BGE chips (5761, 5784 and 57780). These chips uses new BGE_PCI_PRODID_ASICREV register to store the chip identifier and its revision. - Add new grouping macro for 7575+ chips (BGE_IS_5755_PLUS). - Add IDs for Fujitsu-branded Broadcom adapters. PR: kern/127587 Tested by: Thomas Quinot <thomas@quinot.org> (BCM7561 A0) MFC after: 2 weeks Obtained from: OpenBSD
2009-10-07 13:12:43 +00:00
if (BGE_IS_5755_PLUS(sc))
printf(" - 5755 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_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");
- Fix BCM5754 support found in Dell PowerEdge SC440. - Move some PHY bug detections from brgphy.c to if_bge.c. - Do not penalize working PHYs. - Re-arrange bge_flags roughly by their categories. - Fix minor style(9) nits. PR: kern/107257 Obtained from: OpenBSD Tested by: Mike Hibler <mike at flux dot utah dot edu>
2007-01-15 21:43:43 +00:00
if (sc->bge_flags & BGE_FLAG_NO_3LED)
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(" - 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
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
static int
bge_get_eaddr_fw(struct bge_softc *sc, uint8_t ether_addr[])
{
if (sc->bge_flags & BGE_FLAG_EADDR)
return (1);
#ifdef __sparc64__
OF_getetheraddr(sc->bge_dev, ether_addr);
return (0);
#endif
return (1);
}
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
static int
bge_get_eaddr_mem(struct bge_softc *sc, uint8_t ether_addr[])
{
uint32_t mac_addr;
mac_addr = bge_readmem_ind(sc, 0x0c14);
if ((mac_addr >> 16) == 0x484b) {
ether_addr[0] = (uint8_t)(mac_addr >> 8);
ether_addr[1] = (uint8_t)mac_addr;
mac_addr = bge_readmem_ind(sc, 0x0c18);
ether_addr[2] = (uint8_t)(mac_addr >> 24);
ether_addr[3] = (uint8_t)(mac_addr >> 16);
ether_addr[4] = (uint8_t)(mac_addr >> 8);
ether_addr[5] = (uint8_t)mac_addr;
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
return (0);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
}
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
return (1);
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
}
static int
bge_get_eaddr_nvram(struct bge_softc *sc, uint8_t ether_addr[])
{
int mac_offset = BGE_EE_MAC_OFFSET;
if (sc->bge_asicrev == BGE_ASICREV_BCM5906)
mac_offset = BGE_EE_MAC_OFFSET_5906;
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
return (bge_read_nvram(sc, ether_addr, mac_offset + 2,
ETHER_ADDR_LEN));
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
}
static int
bge_get_eaddr_eeprom(struct bge_softc *sc, uint8_t ether_addr[])
{
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
if (sc->bge_asicrev == BGE_ASICREV_BCM5906)
return (1);
return (bge_read_eeprom(sc, ether_addr, BGE_EE_MAC_OFFSET + 2,
ETHER_ADDR_LEN));
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
}
static int
bge_get_eaddr(struct bge_softc *sc, uint8_t eaddr[])
{
static const bge_eaddr_fcn_t bge_eaddr_funcs[] = {
/* NOTE: Order is critical */
Improve the integration of BCM5906[M] support: - Rename BGE_FLAG_EEPROM to BGE_FLAG_EADDR to underline it's absence means "there's no chip containing an Ethernet address fitted to the BGE chip so we have to get it from the firmware instead" rather than "there's no EEPROM, but maybe NVRAM or something else". - Don't treat BCM5906[M] generally like chips w/o BGE_FLAG_EADDR set, just in the two cases really necessary. This gets us line with the original patch for DragonFlyBSD. - For sparc64 restore the intended behavior of obtaining the Ethernet address from the firmware in case BGE_FLAG_EADDR is not set, even for BCM5906[M]. - Fix some style(9) bugs introduced with rev. 1.208 of if_bge.c Approved by: jhb Additional testing by: Thomas Nystroem (BCM5906)
2008-05-14 21:00:27 +00:00
bge_get_eaddr_fw,
Add support for the BCM5906[M] adapters. These adapters only support 10/100 operation and place the mailbox registers at a different offset. They also do not have an EEPROM, so the MAC address must be read from NVRAM instead. MFC after: 1 month PR: kern/118975 Submitted by: benjsc, Thomas Nyström thn at saeab dot se Submitted by: sephe (original patch for DragonflyBSD)
2008-04-29 19:47:13 +00:00
bge_get_eaddr_mem,
bge_get_eaddr_nvram,
bge_get_eaddr_eeprom,
NULL
};
const bge_eaddr_fcn_t *func;
for (func = bge_eaddr_funcs; *func != NULL; ++func) {
if ((*func)(sc, eaddr) == 0)
break;
}
return (*func == NULL ? ENXIO : 0);
}
Reference in New Issue Copy Permalink